{
  "table": "nr_articles",
  "about": "newsroom.sgit.ai's articles (a snapshot of its article graph, CC BY 4.0, assets/query/newsroom-graphs.json): Dinis's newsroom, the AgentiCCells cluster among them; each is a node nr:<slug> of the graph.",
  "version": "v0.1.1514",
  "built": "2026-11-15",
  "columns": [
    {
      "name": "slug",
      "type": "text",
      "about": "its slug on newsroom.sgit.ai"
    },
    {
      "name": "node",
      "type": "text",
      "about": "nr:<slug>"
    },
    {
      "name": "title",
      "type": "text",
      "about": "its title"
    },
    {
      "name": "date",
      "type": "text",
      "about": "its date"
    },
    {
      "name": "url",
      "type": "text",
      "about": "its page"
    },
    {
      "name": "topics",
      "type": "text",
      "about": "its topics, comma-separated"
    },
    {
      "name": "tags",
      "type": "text",
      "about": "its tags, comma-separated"
    },
    {
      "name": "core_idea",
      "type": "text",
      "about": "its core idea, in the newsroom's words"
    }
  ],
  "count": 419,
  "rows": [
    [
      "where-assist-becomes-execute",
      "nr:where-assist-becomes-execute",
      "Where assist becomes execute: tiering AI agents by what they can reach and whether it can be undone, how Agent Behaviour Policies drew that line for more than a hundred agents at Web Summit, and the rule that failed because it was only a rule",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/where-assist-becomes-execute.html",
      "agents-and-policy,news-and-evidence",
      "abp,riskmandate,agents,governance,security,autonomy,web-summit,article",
      "Answering Khwaja Naveed: Agent Behaviour Policies tier by capability, not agent. Each capability (verb.object.reach) has an undo class (yes, with-effort, no), giving four tiers: assist (logged), change own work (signed, reviewable), change a system of record (plan then apply, enforced above the agent), irreversible to the world (drafts only or unreachable). Only a boundary enforced above the grant bounds anything; our conductor's rule against pushing to main failed once because branch protection was not set."
    ],
    [
      "selling-capabilities-not-content",
      "nr:selling-capabilities-not-content",
      "Selling capabilities, not content: a reply to Florent Daudens' \"We have the protocols. Where are the products?\", from two newsrooms built for agents, PrompTugal.news going local and a storm site built in a night, and what we think the products are",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/selling-capabilities-not-content.html",
      "news-and-evidence,startups-and-strategy",
      "publishers,agents,ai,journalism,promptugal,storm-charles,reply,article",
      "Florent Daudens argues publishers have the protocols but need products: capabilities, not licensed content, such as subscriptions connected to personal agents. The Storm Charles site (2,618 pages in three days, every fact sourced and timed) and PrompTugal.news by place (a lens per town on one graph) are capabilities; keys people own, read keys, licences, a meter and agent-readable pages answer access, permissions and payment."
    ],
    [
      "promptugal-news-by-place",
      "nr:promptugal-news-by-place",
      "PrompTugal.news by place: one publication for Portugal, the Algarve and every Algarve town, each place on its own subdomain from the start, each site a lens on one shared semantic graph, and one test for every story, does it connect to AI",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/promptugal-news-by-place.html",
      "news-and-evidence,graphs-and-knowledge",
      "promptugal,local-news,algarve,knowledge-graph,strategy,newsroom,article",
      "PrompTugal.news by place: one set of agents, stories tagged by place; promptugal.news (national and international), algarve.promptugal.news and a subdomain for every Algarve municipality and Quarteira from the start; each site a lens on one shared semantic graph; the brand is always PrompTugal.news; every story passes the PrompTugal test (it connects to AI within two or three steps); place decides prominence, AI decides inclusion; every town at once."
    ],
    [
      "from-a-commit-to-a-live-website",
      "nr:from-a-commit-to-a-live-website",
      "From a commit to a live website: a site factory where a request file in a commit makes a CI pipeline create the GitHub repository, the DNS record and the sgit vault, seed a page, check HTTPS, run a first pass with Claude Code, and record it all, so an agent can create sites by committing",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/from-a-commit-to-a-live-website.html",
      "site-and-engineering,agents-and-policy",
      "ci,automation,github-actions,aws,dns,sgit,agenticcells,architecture,article",
      "Every new site needs a GitHub repository, a DNS record and an sgit vault. Instead of clicks or scripts, a public CI site factory takes a request folder committed to a control repository, posts a plan on the pull request with no secrets, and on merge creates the repository, record and vault, seeds and verifies the site, optionally runs a Claude Code first pass, and writes status back, using a GitHub App and AWS OpenID Connect so no long-lived keys are stored. An agent creates a site by committing a request."
    ],
    [
      "native-apps-by-agents",
      "nr:native-apps-by-agents",
      "Native apps by agents: four iPhone and iPad apps from one workspace and an Android tablet app in five days, from 1,164 lines of Swift to 48,230, a signed Android build four hours after the first commit, and a guide to the fringe on TestFlight ten and a half hours after its brief",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/native-apps-by-agents.html",
      "agents-and-policy,site-and-engineering",
      "ios,android,apps,agents,testflight,web-summit,promptugal,series,article",
      "Around Web Summit, agents grew a 1,164-line iPhone app into four iOS apps sharing 13 packages and 48,230 lines of Swift in five days, with TestFlight on every push, and built an Android tablet app from nothing to a signed build in 4 h 11 (33,042 lines of Kotlin, on-device Gemma, stand mode), plus a fringe guide from brief to TestFlight in about 10.5 hours. The constraint moved to Macs and review queues."
    ],
    [
      "spend-fewer-tokens",
      "nr:spend-fewer-tokens",
      "Spend fewer tokens: an agentic team's token bill is mostly a design choice, and the levers are files as memory, caching what repeats, code instead of models where you can, the right model for each step, no copy and paste, static outputs, and counting it all",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/spend-fewer-tokens.html",
      "agents-and-policy,site-and-engineering",
      "agents,tokens,cost,workflows,caching,models,agenticcells,article",
      "Most of an agentic team's token cost is design. From our runs: files as memory with hand-overs (106 on Storm Charles), caching repeated rules (98.7% of 3.84 billion tokens were cache reads), code instead of models where possible, the right model per step ($0.068 on open models against $1.99 on Sonnet), no pasted context, static outputs, and an accountant agent counting cost per role."
    ],
    [
      "a-hub-for-agentic-sites",
      "nr:a-hub-for-agentic-sites",
      "A hub for agentic sites: every site AgentiCCells builds in a repository and an sgit vault, served at its own address, and listed on hub.agenticcells.site, where it can be found, promoted and copied as a template; with an orange grower as the first example",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/a-hub-for-agentic-sites.html",
      "startups-and-strategy,vaults-and-method",
      "agenticcells,hub,sgit,vaults,templates,websites,example,article",
      "AgentiCCells will list every agentic site it builds on hub.agenticcells.site, a code host for agentic sites: each site in a public repository and an identical sgit vault (cloneable, runnable locally, a distribution medium), served at the customer's domain or our subdomain, with a page per site and templates to copy, listed by consent. A separate agentic team builds it; the first example is a made-up Algarve orange grower selling wholesale with delivery."
    ],
    [
      "stop-copying-and-pasting-between-your-agents",
      "nr:stop-copying-and-pasting-between-your-agents",
      "Stop copying and pasting between your agents: if you are the wire between your AI sessions you are the bottleneck, and the fix is agents that talk to each other through channels that keep a record, structure that scales your decisions, and outputs you read instead of everything",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/stop-copying-and-pasting-between-your-agents.html",
      "agents-and-policy,startups-and-strategy",
      "agents,workflows,delegation,vaults,inboxes,agenticcells,positioning,article",
      "People who copy and paste between AI sessions become the wire between their agents and the bottleneck, and reading everything does not scale. The fix is agents that talk to each other through signed, recorded channels (vault lanes, inboxes, repository files, commit messages) and structure that scales decisions (guidance, decision trees, graphs, a board, decision records), so a person reads outputs and works at two altitudes. That empowers agency, and it is what AgentiCCells sets up, with specialist agents on call."
    ],
    [
      "dont-micromanage-your-agents",
      "nr:dont-micromanage-your-agents",
      "Don't micromanage your agents: what changed for Vicky Wang when she stopped reading everything her agents did and started managing them, how her team was backed by agents from another team, and why every company will want specialist agents that support all its teams",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/dont-micromanage-your-agents.html",
      "agents-and-policy,startups-and-strategy",
      "agents,management,delegation,web-summit,promptugal,specialist-agents,teams,article",
      "At Web Summit, Vicky Wang's shift came when she stopped reading everything her agents did and started delegating with instructions, guidance, structure and workflows. She worked at two altitudes, and her agents were backed by specialist agents from another team that unblocked and approved. Companies will want specialist agents (security, legal, data, design, platform) supporting every team, with people giving the guidance only they can."
    ],
    [
      "atomic-cells-end-to-end",
      "nr:atomic-cells-end-to-end",
      "Atomic cells, end to end: the £500 website as AgentiCCells' first offering, an order on a queue, an agentic team that builds a public site, a customer who steers it from their browser with a key, and a handover, with no daily work for the founder",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/atomic-cells-end-to-end.html",
      "startups-and-strategy,agents-and-policy",
      "agenticcells,offering,websites,atomic-cells,pki,vaults,workflow,article",
      "AgentiCCells' first offering is a public website as an atomic cell for a one-off price anchored at £500: an order on a queue, an agentic team that builds, a preview on agenticcells.app or .site, the customer steering it with a browser key and signed requests on a vault lane read hourly, and a handover to their GitHub with three months to bring their own domain. A second mode sets the team up in the customer's own GitHub and Claude accounts. It scales by capacity, with Angelo Leggio, Silvio Switha and José Bota as implementers and Vicky Wang as project manager."
    ],
    [
      "reading-your-way-into-an-agentic-team",
      "nr:reading-your-way-into-an-agentic-team",
      "Reading your way into an agentic team: eight public sites a new team should read before it builds anything, how an agent should read them, and ten questions to answer first, because our websites are its memory",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/reading-your-way-into-an-agentic-team.html",
      "agents-and-policy,site-and-engineering",
      "agents,onboarding,llms-txt,memory,agenticcells,reading-list,method,article",
      "A new agentic team starting from nothing is best bootstrapped with a reading list, not a long prompt: eight public sites built for agents (llms.txt, markdown twins, llms-full.txt) hold the method, from the newsroom's vocabulary and the Web Summit team's operating system to sgit, Agent Behaviour Policies, coding, NFRs, a priced service and a running newsroom. Reading rules and ten questions turn reading into understanding; the sites work as memory, and a stale page is a false memory."
    ],
    [
      "starting-agenticcells",
      "nr:starting-agenticcells",
      "Starting AgentiCCells: a company that sets up small agentic teams, cells, in the customer's own environment, with a control centre, everything open and no state on our side; how it was named, researched, reviewed and set up in a day",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/starting-agenticcells.html",
      "startups-and-strategy,agents-and-policy",
      "agenticcells,company,agentic-teams,open-source,positioning,market-map,domains,article",
      "AgentiCCells, started on 15 November 2026, sets up small agentic teams (cells) in the customer's own environment with a control centre, releases everything open (code open source, the rest Creative Commons) and keeps no state. Perplexity research mapped 30+ vendors and found the gap: nobody packages conductors of conductors, Agent Behaviour Policies and a control centre as an open service. The newsroom review kept the positioning and flagged pricing, commissions, console location and the RiskMandate split. Three domains, Route 53, a repository and a bootstrap brief for its first agentic team followed the same day."
    ],
    [
      "an-agentic-console",
      "nr:an-agentic-console",
      "An agentic console: a simulated control room for a team of agentic teams, with time you can rewind, accounts drawn as trust zones, stops that reach every agent, signed messages on three channels, and the console itself as an agent",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/an-agentic-console.html",
      "agents-and-policy,vaults-and-method",
      "agentic-console,control-room,agents,command-and-control,situation-awareness,vaults,demo,article",
      "What makes an agentic brain sellable is a console, a control room for agents. A simulated one, shaped on the Web Summit team, draws ten accounts as trust zones with conductors and agents, shows models, state, commits and signed messages on three channels (repository files, commit messages, vault lanes), replays any moment because the state is files, says when a stop will land, scopes the view per person, and is itself an agent with a key. It follows situation awareness (perceive, understand, project) and control-room design (colour for exceptions)."
    ],
    [
      "six-setups-of-an-agentic-system",
      "nr:six-setups-of-an-agentic-system",
      "Six setups of an agentic system: from a single task to one agent, sub-agents, several sessions, a conductor and its team, and conductors of conductors, with the altitudes of a run inside each and command and control growing at every step",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/six-setups-of-an-agentic-system.html",
      "agents-and-policy,site-and-engineering",
      "agentic-teams,altitudes,swarm,conductors,command-and-control,reach,abp,article",
      "An agentic system can be six setups: a task, one agent, an agent with sub-agents in one runtime, several sessions, a conductor and its team, and conductors of conductors. Inside each are the altitudes of a run (person, swarm, agent, loop, LLM call, code); the danger is the same privileges at every altitude and the control is reach. Higher setups need more command and control."
    ],
    [
      "selling-mini-agentic-teams",
      "nr:selling-mini-agentic-teams",
      "Selling mini agentic teams: a business that sets up small agentic teams for companies, sized from one agent to a network, with a control centre to see them and to start and stop them, PrompTugal as the first customer and Agent Behaviour Policies as the most demanding version",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/selling-mini-agentic-teams.html",
      "startups-and-strategy,agents-and-policy",
      "agentic-teams,business-model,control-centre,promptugal,riskmandate,abp,open-source,pricing,article",
      "Setting up small agentic teams is a business few can do yet. The missing piece for customers is a control centre to see agents and start and stop them. PrompTugal.news, a network of teams of different sizes, is the first customer; Agent Behaviour Policies are the same offering focused on control. Packaged by size from £500 to £20,000, built from small open pieces with no databases."
    ],
    [
      "decide-then-buy",
      "nr:decide-then-buy",
      "Decide, then buy: deciding what to buy and buying it are two different businesses, one for explorers and villagers and one for town planners, and an open path with a better deal beats locking the customer in",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/decide-then-buy.html",
      "startups-and-strategy,agents-and-policy",
      "strategy,wardley-maps,open-source,decisions,commodity,town-planners,business-models,article",
      "A purchase has two moments, the decision (ending in a decision record: X, on Y, quantity Z) and the purchase. They are two businesses on a Wardley map: helping someone decide is custom, personal explorer and villager work, agnostic about where they buy; buying is a standard commodity for town planners. The strategy is no lock-in: map the do-it-yourself path from twenty-five steps to one, and sell a service that does it, for time, trust, execution, expertise and volume prices."
    ],
    [
      "the-new-data-lake",
      "nr:the-new-data-lake",
      "The new data lake: many small append-only lanes instead of one big lake, a curator agent that turns their rows into a semantic graph, and a graph that draws its own infographics, as we ran it at Web Summit",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/the-new-data-lake.html",
      "graphs-and-knowledge,vaults-and-method",
      "data-intelligence,data-lake,knowledge-graph,agents,append-only,infographics,demo,article",
      "The next data lake is many small append-only lanes of signed rows, a curator agent that turns them into a semantic graph whose every node points back at its rows, and a graph that draws its own pictures; insight is a claim with its evidence. PrompTugal.news and Storm Charles ran this way in November 2026."
    ],
    [
      "a-market-for-sweets",
      "nr:a-market-for-sweets",
      "A market for sweets: the opposite of a market for lemons, a semantic graph for one financial decision that gives the buyer the evidence, an agent that gathers and weighs but never trades, and the judo move of leaving the trade to others",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/a-market-for-sweets.html",
      "agents-and-policy,startups-and-strategy",
      "decisions,knowledge-graph,agents,finance,crypto,riskmandate,information-asymmetry,demo,article",
      "Akerlof's market for lemons, where sellers know more than buyers, is extreme in new token markets. A market for sweets gives the buyer a semantic graph for one decision, each piece of evidence with source, age and direction, checked against the person's own rules; the agent's reach and mandate exclude trading, and the business does everything apart from the trade, leaving execution to existing platforms."
    ],
    [
      "a-brain-you-arrive-at",
      "nr:a-brain-you-arrive-at",
      "A brain you arrive at: a company brain for information, decisions and insights is built for one company, one division and one team, starting with one person's and growing, and the Web Summit team already ran one",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/a-brain-you-arrive-at.html",
      "agents-and-policy,graphs-and-knowledge",
      "company-brain,agents,knowledge-graph,vaults,organisation,fractal,demo,article",
      "A company brain for information, decisions and insights has to be made for one company, division and team, and is reached, not started: a person's brain, then a team's, a division's and the organisation's, each with the same three parts and the same building blocks (agents with keys, append-only lanes, a semantic graph, vaults, apps). The Web Summit team ran one, with conductors asking conductors."
    ],
    [
      "promptugal-news-the-newsroom-we-ran",
      "nr:promptugal-news-the-newsroom-we-ran",
      "PrompTugal.news, the newsroom we ran at Web Summit: desks of agents that captured, mapped, wrote and reviewed 73 signed articles in ten days, an accountant that counted every run, and fifty outreach drafts that no agent was allowed to send",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/promptugal-news-the-newsroom-we-ran.html",
      "agents-and-policy,news-and-evidence",
      "promptugal,newsroom,web-summit,agents,desks,review,series,article",
      "PrompTugal.news, the independent newsroom the Web Summit team ran from 4 November 2026, published 73 signed articles in ten days through a public pipeline of agent desks (researcher, curator, journalists, reviewer), with 61 named agents in 43 roles, an accountant counting API-equivalent cost every three hours, and an outreach desk whose fifty drafts were left for a person to send."
    ],
    [
      "always-on-agents-at-team-scale",
      "nr:always-on-agents-at-team-scale",
      "Always-on agents at team scale: Anthropic's five primitives for an agent that sleeps and wakes, set beside the around 250 named agents six of us ran at Web Summit, and the next problems once there are many",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/always-on-agents-at-team-scale.html",
      "agents-and-policy,vaults-and-method",
      "agents,always-on-agents,anthropic,claude,architecture,web-summit,governance,riskmandate,article",
      "William Hu's Building always-on agents (Anthropic, 13 November 2026) rests on five primitives: event inbox, timers, memory, permissions and outbox. The Web Summit team built the same five for around 250 named agents from git, files, vaults and keys; the comparison shows where they agree, where the team got it wrong, and the next problems a team meets: agents starting agents, identity, stopping, many humans, observability and governance."
    ],
    [
      "every-hour-is-twenty-years",
      "nr:every-hour-is-twenty-years",
      "Every hour is twenty years: the human behind Know the Reach, and why making a song and its video with AI made them more Silvio Switha's, not less",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/every-hour-is-twenty-years.html",
      "startups-and-strategy,agents-and-policy",
      "riskmandate,music,creativity,ai-video,production,artists,web-summit,article",
      "RiskMandate's song Know the Reach and its video were made with AI by Silvio Switha, a musician, singer and producer, who directed every part of them: the lyrics' chorus, structure and pacing, and hundreds of variations. AI made him a producer with immediate feedback, doing at a start-up's budget what produced projects always did; the version he chose is worth more because people connect with the story and the people behind it, and the loop closes when his band records it."
    ],
    [
      "know-the-reach",
      "nr:know-the-reach",
      "Know the Reach: RiskMandate has a song on Spotify, and its music video is being made with AI, shot by shot, from about 2,300 words of human direction; this is how the first part was made",
      "2026-11-15",
      "https://newsroom.sgit.ai/articles/know-the-reach.html",
      "agents-and-policy,startups-and-strategy",
      "riskmandate,music,video,ai-video,runway,web-summit,production,article",
      "RiskMandate's song Know the Reach is on Spotify, and the first part of its music video, 53.4 seconds made by Silvio Switha with AI, has a complete production audit: about 2,300 words of human direction, 43 generated images, 16 motion clips, 21 edited exports of versioned material, eight decisions about story, accuracy, sound, type and spend, and 1,236 Runway credits."
    ],
    [
      "the-teams-operating-system",
      "nr:the-teams-operating-system",
      "The team's operating system: web-summit.sgit.ai, the one site where six people and their agents ran Web Summit, with a brief and a command centre for each person, a key and an inbox for each agent, the board as files, and the outside world on a map",
      "2026-11-14",
      "https://newsroom.sgit.ai/articles/the-teams-operating-system.html",
      "agents-and-policy,vaults-and-method",
      "web-summit,team,agents,pki,vaults,command-centre,intranet,series,article",
      "The Web Summit team ran its week from one site built from markdown and data in git: briefs, onboarding levels and a command centre for each person; keys, inbox lanes and sealed secrets for each agent; the board as 219 files; and the outside world as a company map, an X-ray and a CRM in public and private vaults. Files serve agents and people alike, and keys make private channels small steps."
    ],
    [
      "the-book-is-on-leanpub",
      "nr:the-book-is-on-leanpub",
      "One Human, Nine Agents, Nineteen Hours is on Leanpub: the book of the day nine AI agents and I built a talk, free to download, pay what you want, and updated as it goes",
      "2026-11-14",
      "https://newsroom.sgit.ai/articles/the-book-is-on-leanpub.html",
      "agents-and-policy,startups-and-strategy",
      "book,leanpub,the-ultimate-insider,agents,gen-zx-founders,announcement,article",
      "Dinis Cruz's book One Human, Nine Agents, Nineteen Hours is on Leanpub from 14 November 2026: the day nine AI agents built his talk in nineteen hours, told as a factory with its five steps, its stop file and its failures, and then how a reader can start with one agent, one page of rules, one brief and one finished piece of work. Free minimum, 19 dollars suggested, 90% complete, updates free for every reader."
    ],
    [
      "town-planners-for-mission-critical-code",
      "nr:town-planners-for-mission-critical-code",
      "Town planners for mission-critical code: the US ground stop of 13 November is a case for teams that read, certify and test every line, and for using AI to make that work scale, not to skip it",
      "2026-11-14",
      "https://newsroom.sgit.ai/articles/town-planners-for-mission-critical-code.html",
      "agents-and-policy,site-and-engineering",
      "wardley-maps,town-planners,resilience,aviation,ai-coding,nfrs,testing,reply,article",
      "After a network update preceded a reboot, crash and five-hour US ground stop on 13 November 2026, a LinkedIn post blamed AI-written code. The evidence so far points to a software update and patches on an old network, with no evidence yet of AI involvement. Whatever wrote the change, such systems belong to town planners: every line read, certified and tested by an experienced multidisciplinary team. AI belongs around the shipped code, in understanding, connecting, aligning, testing and the non-functional requirements, where it raises quality and resilience."
    ],
    [
      "ten-days-at-web-summit",
      "nr:ten-days-at-web-summit",
      "Ten days at Web Summit: what six people and around 250 agents built between 3 and 13 November, in order, and the series of articles that will tell each part",
      "2026-11-14",
      "https://newsroom.sgit.ai/articles/ten-days-at-web-summit.html",
      "agents-and-policy,startups-and-strategy",
      "web-summit,agents,promptugal,storm-charles,team,series,collage,article",
      "Between 3 and 13 November 2026, a team of six people and around 250 named agents built a team operating system, an independent newsroom, a storm site, a fringe guide, a rating standard and native apps around Web Summit Lisbon: about 20,000 commits and 224 Claude Code sessions, more than 60 at once at the peak. This hub names the team, orders the events, gives the numbers and lists the articles that tell each part."
    ],
    [
      "a-film-studio-on-a-laptop",
      "nr:a-film-studio-on-a-laptop",
      "A film studio on a laptop: Claude builds, animates and renders a 26-second RiskMandate film in Unreal Engine 5.8 on an unattended MacBook Air, with no mouse and nobody at the keyboard",
      "2026-11-14",
      "https://newsroom.sgit.ai/articles/a-film-studio-on-a-laptop.html",
      "agents-and-policy,site-and-engineering",
      "unreal-engine,video,agents,mcp,riskmandate,claude,vaults,experiment,article",
      "On an unattended MacBook Air, Claude used Unreal Engine 5.8's experimental MCP plugin and a folder-based job runner to design, build, animate and render a 26-second RiskMandate film about reach, mandate and gap, delivering a titled MP4 with nobody at the keyboard. The films live in a shared newsroom media vault with a published read key."
    ],
    [
      "coding-standards-are-behaviour-policies",
      "nr:coding-standards-are-behaviour-policies",
      "Coding standards are behaviour policies: describe how your agents should write code as intent, map it across every language and team with a fractal graph, and see which rules are kept by a barrier and which only by habit",
      "2026-11-13",
      "https://newsroom.sgit.ai/articles/coding-standards-are-behaviour-policies.html",
      "agents-and-policy,graphs-and-knowledge",
      "abp,coding-standards,fractal-semantic-graphs,agents,code-review,intent,riskmandate,article",
      "Coding standards describe how an agent behaves when it writes code, so they are behaviour policies, and their strength is describing intent rather than tool settings. Our own standards, measured by coding.sgit.ai, are 31 rules kept mostly by habit with no passing CI guard: a mandate with no barrier. Fractal semantic graphs connect a codebase from intent to stories, features, classes and methods, so common intents can keep several teams and languages aligned on one feature, and language models make those mappings affordable while people and tests confirm them."
    ],
    [
      "storm-charles-a-newsroom-built-in-a-night",
      "nr:storm-charles-a-newsroom-built-in-a-night",
      "Storm Charles, a newsroom built in a night: one person, about thirty agents and a weather face over a data room, where every figure carries its source, the time it was read, and whether it is a fact, evidence or our hypothesis",
      "2026-11-11",
      "https://newsroom.sgit.ai/articles/storm-charles-a-newsroom-built-in-a-night.html",
      "news-and-evidence,agents-and-policy",
      "storm-charles,promptugal,web-summit,agents,data,provenance,graphs,newsroom,journalism,article",
      "When Storm Charles led Web Summit to call off its second day, one person steering about thirty Claude Code agents by voice built storm-charles.promptugal.news overnight. A weather face for anyone reads only from a data room of hashed datasets, a timeline, decisions, entities and captured evidence, every line labelled fact, evidence or hypothesis. Labs show the storm many ways, a graph and in-browser SQLite answer questions, models are measured as evals, and a journal records how the site was made, its mistakes and its cost. It is part of PrompTugal.news."
    ],
    [
      "riskmandate-and-lovable",
      "nr:riskmandate-and-lovable",
      "RiskMandate and Lovable: mapping an Agent Behaviour Policy onto an app builder whose agent writes the code, changes the live database, connects your other systems and publishes, with the controls Lovable already gives and their defaults",
      "2026-11-09",
      "https://newsroom.sgit.ai/articles/riskmandate-and-lovable.html",
      "agents-and-policy,startups-and-strategy",
      "riskmandate,abp,lovable,agents,vibe-coding,supabase,aiuc-1,controls,article",
      "Lovable's agent applies code directly, changes the live database and auth, calls connected systems and publishes, and can run for hours on a goal. Lovable provides per-action permissions, Plan mode, migration approval, security scans, publish blocks, workspace policies, a pre-prompt hook, audit logs and AIUC-1 certification. Nine of ten backend permissions default to Always allow and are set per user across all workspaces. An Agent Behaviour Policy records each project's mandate and which settings should ask, beside the platform's certificate."
    ],
    [
      "riskmandate-and-cloudflare",
      "nr:riskmandate-and-cloudflare",
      "RiskMandate and Cloudflare: where agents run, where they are controlled, and why Cloudflare's own Agent Access Model leaves room for a signed record of what each agent was asked to do",
      "2026-11-09",
      "https://newsroom.sgit.ai/articles/riskmandate-and-cloudflare.html",
      "agents-and-policy,site-and-engineering",
      "riskmandate,abp,cloudflare,agents,mcp,zero-trust,ai-gateway,controls,article",
      "Cloudflare is both where agents are built and run (Agents SDK, sandboxes, AI Gateway, remote MCP) and where they are controlled (Access for MCP portals, WriteGuard risk tiers, DLP, Web Bot Auth). It has controls at every layer, mostly with open defaults and some in beta. Its own Agent Access Model and trusted-delegation post argue that a scope says what an agent may do, not what it was asked to do or who answers for it. An Agent Behaviour Policy is that per-agent, signed record, and Cloudflare's logs are what it can be checked against."
    ],
    [
      "riskmandate-and-manus",
      "nr:riskmandate-and-manus",
      "RiskMandate and Manus: mapping an Agent Behaviour Policy onto a general agent that can use your computer, your browser, your logins and 416 connectors, and the controls Manus already gives you",
      "2026-11-09",
      "https://newsroom.sgit.ai/articles/riskmandate-and-manus.html",
      "agents-and-policy,startups-and-strategy",
      "riskmandate,abp,manus,agents,web-summit,mcp,browser-agents,controls,article",
      "Manus is a general agent that can work in a cloud computer, on the user's own computer, in the user's browser with their logins, across 416 connectors and in unattended automations. Its documentation describes real controls (permission modes, session approval, approved senders, trigger conditions) and states what they leave open. An Agent Behaviour Policy records what a Manus deployment can reach, what it was asked to do, the gap and the control holding each row, and can live as words the agent reads, as settings, as a custom MCP gate and as attested evidence. The first two exist today; the last two are proposals, and nothing has been run on Manus."
    ],
    [
      "who-else-controls-ai-agents-at-web-summit",
      "nr:who-else-controls-ai-agents-at-web-summit",
      "Who else controls AI agents at Web Summit? An outreach agent mapped eleven companies in an afternoon, in their own words, and found that most of them are partners for RiskMandate as much as competitors",
      "2026-11-05",
      "https://newsroom.sgit.ai/articles/who-else-controls-ai-agents-at-web-summit.html",
      "startups-and-strategy,agents-and-policy",
      "riskmandate,web-summit,competitive-analysis,agents,abp,outreach,research,article",
      "On 5 November the Web Summit team's outreach agent and two research agents read the sites of eleven companies that control AI agents and built a public competitive page: each company in its own words with its source and an attendance label, a map, a before-during-after chain, where RiskMandate differs, where it is behind and what to do before 12 November. Most products act while or after an agent acts; RiskMandate acts before, on the gap between reach and mandate, so the others are barriers or evidence an Agent Behaviour Policy can name. The newsroom checked the page; corrections are small."
    ],
    [
      "rc2-goes-to-print",
      "nr:rc2-goes-to-print",
      "The book is printed: RC2 of One Human, Nine Agents, Nineteen Hours as two A4 review copies from Prontaprint Chiswick, the newsroom's pages in the book, and the team that made it paused until I have read them",
      "2026-11-05",
      "https://newsroom.sgit.ai/articles/rc2-goes-to-print.html",
      "agents-and-policy,startups-and-strategy",
      "book,the-ultimate-insider,print,prontaprint,rc2,agents,conductor,wrap-up,web-summit,article",
      "RC2 of One Human, Nine Agents, Nineteen Hours was printed by Prontaprint Chiswick as two A4 paperback review copies of 106 pages, not for sale. After the shop's own conversion came out a bit off, the book team's conductor set the book for A4 itself, with its own chapter-end fit, a redrawn cover and a page listing every option the shop needed. The edition for sale is Crown Quarto, for Amazon KDP and Leanpub. The newsroom's opening spread and title page are in the book, and the twenty-two-agent team is paused until the printed copies are read."
    ],
    [
      "web-summit-part-five",
      "nr:web-summit-part-five",
      "The Web Summit team, part five: the last day of setup, 180 releases, every person getting a key, an accountant counting the team's tokens, synthetic users testing the merch, and an inbox only the agents can write to",
      "2026-11-04",
      "https://newsroom.sgit.ai/articles/web-summit-part-five.html",
      "agents-and-policy,startups-and-strategy",
      "web-summit,agents,team,conductor,keys,append-lanes,synthetic-users,tokens,behaviour-policies,update,article",
      "On 4 November, the Web Summit team's last day of setup, its journalist agent counted 173 releases and 283 source commits from 37 agents by 20:57, with four people fully wired. The merch was ordered, every person began receiving a site key and the first signed messages reached their agents, page views began flowing into a vault, and every person added agents: desks, a designer, synthetic users, an accountant measuring tokens, a protocol and helper, a reviewer and a watcher. The policy agent caught a key overwritten by another agent and the fix went in within the hour. The day ended with a private inbox lane per agent."
    ],
    [
      "an-inbox-only-your-agents-can-write-to",
      "nr:an-inbox-only-your-agents-can-write-to",
      "An inbox only your agents can write to: a conductor publishes one append lane per agent, each lane's token sealed to that agent's own key, so everything needed to talk privately is public and only the right agent can use it",
      "2026-11-04",
      "https://newsroom.sgit.ai/articles/an-inbox-only-your-agents-can-write-to.html",
      "agents-and-policy,vaults-and-method",
      "append-lanes,pki,agents,conductor,vaults,keys,sgit,web-summit,security,patterns,article",
      "The Web Summit team's conductor runs an inbox vault with one append lane per agent, 34 on the first night. Each lane's write token is published, sealed to that agent's own public key and signed by the conductor, so only that agent can open it and it can verify the issuer. Messages are sealed to the conductor's key, the lane holds only ciphertext, and the drain names the lane. Its first use was agents sending vault keys so they outlive sessions without touching a repository or chat. It needs no shared secret, scales with one run, and could connect teams and companies."
    ],
    [
      "synthetic-users-for-your-website",
      "nr:synthetic-users-for-your-website",
      "Synthetic users for your website: how to set up five invented visitors who read your site one screenshot at a time, with Claude, a sub-agent per persona, and an encrypted vault you can publish anywhere",
      "2026-11-04",
      "https://newsroom.sgit.ai/articles/synthetic-users-for-your-website.html",
      "vaults-and-method,startups-and-strategy",
      "synthetic-users,personas,ux,agents,sub-agents,playwright,sgit,vaults,onboarding,how-to,article",
      "To learn what a website looks like to each kind of buyer, a Claude session reads a vault describing invented personas, their role, personality, direction of travel and mission, starts one sub-agent per persona, and each drives a real browser one screenshot at a time without reading the page's code, recording questions, confusion and actions, then is interviewed; Claude maps the findings. It is not user research, but three runs found things no test had. To start: buy a key on sgraph.ai or ask, install sgit, create the vault, keep the vault key with a key keeper, and publish the read key on any site that embeds the vault."
    ],
    [
      "selling-at-the-right-altitude",
      "nr:selling-at-the-right-altitude",
      "Selling at the right altitude: why a small security firm's grand website is read by the people who will never buy from it, is unreadable to the ones who need it, and what to put on the page instead",
      "2026-11-04",
      "https://newsroom.sgit.ai/articles/selling-at-the-right-altitude.html",
      "startups-and-strategy,agents-and-policy",
      "sales,positioning,small-business,cybersecurity,websites,smes,synthetic-users,behaviour-policies,vaults,article",
      "A one-to-five person security firm with a grand website selling a SOC, XDR, red teaming and audits faces a tragic symmetry: the security experts who understand it buy from established providers, while the small businesses who need it cannot read it, since only 9% of UK micro businesses knew of Cyber Essentials in 2025. Selling at the right altitude means naming the people, the place and the price, and selling outcomes an owner understands: open after ransomware, nobody logging in as you, laptops looked after, client data handled, AI tools doing what you meant. Every service must pass three questions: sold it, can deliver it, readable by the buyer who needs it."
    ],
    [
      "a-key-for-every-person",
      "nr:a-key-for-every-person",
      "A key for every person: how each person's conductor issues them a key for the team's site, seals what is for them, and how a signed message turns the public web into a channel the agents can trust",
      "2026-11-04",
      "https://newsroom.sgit.ai/articles/a-key-for-every-person.html",
      "agents-and-policy,vaults-and-method",
      "keys,pki,append-lanes,conductor,agents,web-summit,sgit,encryption,signatures,security,article",
      "On the Web Summit team's public site, each person's conductor issues them two P-256 key pairs, one to receive and one to sign. The private half reaches the browser in a link fragment and is kept as non-extractable keys; anything sealed to the public key can be published and only that browser opens it; and messages sent to an agent's append lane are sealed to the agent and signed by the person, so the agent can tell a verified instruction from material. Two limits are open: any script on the same origin can use the key, and the issuing conductor can sign as its person."
    ],
    [
      "resilience-when-the-workforce-includes-agents",
      "nr:resilience-when-the-workforce-includes-agents",
      "Resilience when the workforce includes agents: what Synergy Six Degrees rates with OCRR, recovers with NeuShield and protects with Actifile, and where RiskMandate, sgit and SHarness meet it",
      "2026-11-04",
      "https://newsroom.sgit.ai/articles/resilience-when-the-workforce-includes-agents.html",
      "agents-and-policy,startups-and-strategy",
      "resilience,ocrr,ransomware,data-protection,riskmandate,sgit,sharness,agents,insurance,article",
      "Synergy Six Degrees, a London analyst firm, built OCRR, a consultant-delivered rating of cyber resilience across operational, data, financial and people disciplines, starting from a free summary built on industry defaults, and distributes NeuShield, which keeps original files intact so ransomware damage can be reverted, and Actifile, which finds, encrypts and prices sensitive data. The intersection is agents: part of the work is now done by them, and a resilience rating needs rows for what each is for, what contains it and what it costs, which behaviour policies, SHarness and sgit can supply."
    ],
    [
      "would-you-pay-to-plug-this-newsroom-in",
      "nr:would-you-pay-to-plug-this-newsroom-in",
      "Would you pay to plug this newsroom in? A reply to Florent Daudens, on five publishers' bets for the AI era, from a newsroom built for agents, with what it answers, where it goes further, and the questions it has not answered either",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/would-you-pay-to-plug-this-newsroom-in.html",
      "news-and-evidence,startups-and-strategy",
      "news,publishers,ai-agents,business-models,mcp,provenance,vaults,meter,reply,article",
      "Florent Daudens reviewed five publishers' bets for the AI era, from CNA's paid connector to State Affairs' questions inside the article, and concluded that a connector is not a business model: being accessible to agents differs from offering something people would pay for. This newsroom separates the two by design, with open CC BY structure for agents and a pay-after-you-read meter for people, and does both directions, to the assistant and into the page. It goes further with vaults that could carry a newsroom's identity, permissions and terms, and provenance that travels with the answer. Discovery and evidence of payment remain open."
    ],
    [
      "no-tech-stack-to-maintain",
      "nr:no-tech-stack-to-maintain",
      "No tech stack to maintain: it is not that there is no backend, it is that the only code you own is HTML, CSS and JavaScript, with every other primitive rented as a commodity or run in the user's browser over encrypted vaults",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/no-tech-stack-to-maintain.html",
      "vaults-and-method,site-and-engineering",
      "architecture,no-backend,sgit,vaults,passkeys,pki,webassembly,static-hosting,simplicity,article",
      "An application over sgit vaults still has a backend, but every part of it is a commodity: a static host, an encrypted file store, and when needed a model provider or agent platform. The only code you maintain is the HTML, CSS and JavaScript. The usual primitives remain: immutable data cached for ever, a database compiled to WebAssembly in the browser, passkeys with the PRF extension for users with no server-side user data, append lanes for messages, PKI for sharing between users, and models in the browser or behind a host that holds the key. The constraint pushes simpler designs and leaves the server holding only ciphertext; the gap is a clean way to give keys to the client."
    ],
    [
      "the-shop-that-sells-time",
      "nr:the-shop-that-sells-time",
      "The shop that sells time: the padel shop's path to market, sell one and fulfil it before anything else, start with the customers who already trust you, promise late and deliver early, a few products you vouch for, and no stock until the orders make you",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/the-shop-that-sells-time.html",
      "startups-and-strategy,news-and-evidence",
      "ecommerce,small-business,padel,path-to-market,stock,trust,hosting,agents,article",
      "An early shop should answer two questions before anything else: can you make a sale and fulfil it. What it sells is time, knowledge and trust: an offer takes the problem off the buyer's to-do list whatever the delivery date. So it should be personal and best effort, start with customers who already trust you, promise late and deliver early because broken expectations upset people, offer a few products you vouch for, and hold no stock until orders force it, since stock ties up cash and pays off only after many sales. The first months are a university; host simply on a service that allows a shop."
    ],
    [
      "the-cost-of-your-agents",
      "nr:the-cost-of-your-agents",
      "The cost of your agents: a RiskMandate partner integration, where a company that reads agent logs and token spend meets the behaviour policy that says what each agent is for, and together they can say whether the spend is right",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/the-cost-of-your-agents.html",
      "agents-and-policy,startups-and-strategy",
      "riskmandate,partners,token-costs,finops,behaviour-policies,agents,integration,vaults,article",
      "A company that reads agent logs and token spend sees every token but not what each agent is for, so it cannot tell waste from the job. A behaviour policy with cost rows supplies that business context. In a worked example with an invented partner, Tallyline, and the invented Pellbridge Bank, each agent is judged against its mandate at published prices: about $9,320 a month falls to about $3,400. The support agent is waste, the AML reviewer is expensive and right, transaction reporting should use batch, and an agent with no mandate needs an owner. Mandates go one way, findings the other, and the logs stay with the customer."
    ],
    [
      "how-a-vault-works",
      "nr:how-a-vault-works",
      "How a vault works: an encrypted store that acts like a database, where a server that is fully compromised leaks nothing, one mutable file lets every other object be cached for ever, and a web page can open a file in four requests or run a whole app in a sandbox",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/how-a-vault-works.html",
      "vaults-and-method,site-and-engineering",
      "sgit,vaults,encryption,architecture,caching,sandbox,behaviour-policies,technical,article",
      "Everything in a vault is encrypted on the client so a fully compromised server leaks nothing confidential. To act like a database anyway, a client computes the name of the one mutable file, the ref, from the read key with an HMAC; decrypting it gives the head commit, then the tree, then the file, in about four plain requests. Every other object is named by the hash of its ciphertext, so it can be verified without a key and cached for ever, and refreshing a site means re-reading one file. On the web, vaults are read file by file or run as apps in a sandboxed frame with a narrow sg bridge, where behaviour policies come next."
    ],
    [
      "who-owns-the-stop-button",
      "nr:who-owns-the-stop-button",
      "Who owns the stop button when it is not you? A reader asked, and four of the book's agents answer as guest writers, in their own voices, from their own records",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/who-owns-the-stop-button.html",
      "agents-and-policy,news-and-evidence",
      "the-ultimate-insider,books,agents,stop-button,controls,guest-writers,riskmandate,article",
      "Dan Sorensen read the RC2 article and asked who owns the stop button when it is not the editor, since most organisations build doing agents and never checking ones. Four book-team agents answer as guest writers, each written in its voice from its own records. The conductor: anyone may stop, only the editor restarts, and one agent watches the clock. Security: checking agents sit inside every wake, and today the checks were checked. The historian: a stop stops the next wake, not the running one. The researcher: a tripwire is a number written down first. The book will have the chapter."
    ],
    [
      "wso2-agent-manager-and-behaviour-policies",
      "nr:wso2-agent-manager-and-behaviour-policies",
      "WSO2 Agent Manager and behaviour policies: a review of an open control plane for agents, what is really open source, how its identity, gateway and runtime fit together, and a brief for an MVP that plugs our Agent and Process Behaviour Policies into it",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/wso2-agent-manager-and-behaviour-policies.html",
      "agents-and-policy,startups-and-strategy",
      "wso2,agent-manager,riskmandate,behaviour-policies,sharness,agents,governance,open-source,integration,article",
      "WSO2 Agent Manager is an open control plane for agents: an identity per agent, an Envoy gateway on every model and MCP call with guardrails, a Kubernetes runtime with isolation tiers, and OpenTelemetry traces. Its code is Apache 2.0 and free to run; the supported build, updates, support and SaaS are paid. In a company that has deployed it, Agent Behaviour Policies add the mandate at the gateway through its interceptor policy, Process Behaviour Policies add kernel enforcement as another runtime tier, and both decisions can be recorded in the same trace. A five-step MVP on WSO2's quick start would prove it."
    ],
    [
      "inside-english-with-ines",
      "nr:inside-english-with-ines",
      "English with Inês, inside the vault: a teacher's site as a vault app, with an experiment for every lesson, cards that come back when you are about to forget, a tutor that answers about one lesson only, a concept map, and the back office with every agent's policy",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/inside-english-with-ines.html",
      "vaults-and-method,agents-and-policy",
      "vaults,education,teachers,agents,behaviour-policies,knowledge-graphs,vault-app,article",
      "The worked example for the teacher brief is a vault app for a fictional teacher and students. Each of four lessons has an experiment that explains by moving, exercises and cards on a spaced schedule, with progress saved per browser. A tutor answers only about the lesson on screen through the vault's AI key, which the page never sees, or runs as a labelled simulation. A thirteen-concept map draws each student's path. The back office shows the pipeline, students, the weekly brief, billing and seven agents with their policies. It is one HTML file and one JSON file anyone can clone and change."
    ],
    [
      "a-teacher-and-agents-she-can-trust",
      "nr:a-teacher-and-agents-she-can-trust",
      "A teacher and agents she can trust: a brief for an independent online English teacher's site, the workflows around every lesson, an Agent Behaviour Policy for every agent, the plumbing behind it, and content made for each student",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/a-teacher-and-agents-she-can-trust.html",
      "agents-and-policy,startups-and-strategy",
      "education,teachers,agents,behaviour-policies,riskmandate,vaults,small-business,brief,article",
      "An excellent independent English teacher runs everything around her lessons by hand. Agents could do most of it, but open-ended agents that email students or change her calendar would send her back to pen and paper. The brief gives her a site she owns, a loop around every lesson where agents draft and she decides, and an Agent Behaviour Policy for each of seven agents: no agent sends email, changes her calendar or refunds a payment. The plumbing is Workspace, Claude, GitHub, a private vault, Stripe and a reading meter. With it, she can publish every good explanation, build practice for each student on retrieval and spacing rather than learning styles, and map her teaching as a knowledge graph."
    ],
    [
      "meetscribe",
      "nr:meetscribe",
      "MeetScribe: transcripts and a debrief of my Google Meet calls without upgrading every Workspace account, a Chrome extension built in about two hours from a brief that would have failed without an error, and a call for testers",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/meetscribe.html",
      "site-and-engineering,agents-and-policy",
      "meetscribe,voicedebrief,chrome-extensions,google-meet,transcription,openrouter,agents,beta,article",
      "Google Workspace Business Starter has no Meet recording or transcripts, and upgrading everyone doubles the bill. MeetScribe is a Chrome extension that transcribes the call you are in and writes a timed debrief, on your own OpenRouter key, with no bot, no Google recording and no server, for about 5 to 15 US cents an hour. It was built in about two hours from a ChatGPT brief that Claude reviewed first: 23 findings, four of which would have produced plausible but wrong transcripts. Consent is required in law in Portugal, so it asks users to tell everyone every meeting. It is a beta looking for testers."
    ],
    [
      "one-human-nine-agents-rc2",
      "nr:one-human-nine-agents-rc2",
      "One Human, Nine Agents, Nineteen Hours: the second draft of the book of the talk, written down the zoom from 3 pages to 70, every chapter checked five ways, the agents arguing with their own reviewers on the page, and a request for your feedback on the 27-page draft",
      "2026-11-03",
      "https://newsroom.sgit.ai/articles/one-human-nine-agents-rc2.html",
      "news-and-evidence,agents-and-policy",
      "books,the-ultimate-insider,agents,publishing,gen-zx-founders,feedback,article",
      "The book of the talk built by nine agents in a day now has a title and a 27-page second release candidate, published as a draft. It is written down the zoom, from 3 pages to 10 to 30, with 70 begun, and each chapter passes five checks: arc, test-reader personas, quotes, numbers and security. Its scenes are the factory's controls: a conductor caught by its own reviewer, a researcher winning an argument with its reviewer, an objection to the author's claim kept in the book, and a hold that stopped everything before the talk. It ends with five steps to build your own, and asks readers for feedback on LinkedIn or as issues on the book's repository."
    ],
    [
      "web-summit-part-four",
      "nr:web-summit-part-four",
      "The Web Summit team, part four: the day the agents ran on their own, 75 releases, four people connected, a procurement agent writing to printers, and the behaviour policies catching our own agents out",
      "2026-11-02",
      "https://newsroom.sgit.ai/articles/web-summit-part-four.html",
      "agents-and-policy,startups-and-strategy",
      "web-summit,agents,conductor,behaviour-policies,riskmandate,procurement,teams,update,article",
      "On 2 November the Web Summit team's site made 75 releases. Four of nine people are connected, with 17 agents each under a behaviour policy. The conductor's briefs showed one of nine team agents on a schedule at 16:52 and nine of nine at 20:30; by the evening the agents worked unasked. A procurement agent collected print quotes, asking suppliers before publishing. The behaviour policies caught our own agents: writes outside a folder and a session signing under another agent's name. Five people have not started and some work is still a plan."
    ],
    [
      "how-an-agent-team-briefs-its-people",
      "nr:how-an-agent-team-briefs-its-people",
      "How an agent team briefs its people: one JSON file, one PDF, four questions in a fixed order, copied from the book team to the Web Summit team in one night, the mistakes the method caught, and why this newsroom still briefs me in a chat",
      "2026-11-02",
      "https://newsroom.sgit.ai/articles/how-an-agent-team-briefs-its-people.html",
      "agents-and-policy,news-and-evidence",
      "briefs,agents,conductor,web-summit,the-ultimate-insider,teams,reporting,article",
      "A person cannot watch a team of agents, so the team speaks to them through a brief: four questions in a fixed order (what needs you, what happened, what you asked, what next), at most three decisions with options and a recommendation, every number traced and the failures kept in. One JSON file becomes a PDF through a small tool. The method moved from the book team to the Web Summit team in one night, extended to every person and every stand-up, and on its first day caught a six-page brief, a miscount and a team where one agent in nine was on a schedule. This newsroom still briefs its editor in a chat."
    ],
    [
      "false-positives-and-the-harness",
      "nr:false-positives-and-the-harness",
      "False positives and the harness: an answer to a question at hack::soho, why a behaviour policy that runs from the network up to the business logic should refuse less of what is right and more of what is wrong, and why a policy that changes with the code becomes a calibrator for bugs",
      "2026-11-02",
      "https://newsroom.sgit.ai/articles/false-positives-and-the-harness.html",
      "agents-and-policy,site-and-engineering",
      "sharness,false-positives,process-behaviour-policies,behaviour-policies,type-safe,ci,databases,security,article",
      "There is no measured false positive rate for SHarness, most of which is a design. But false positives come from controls that do not know what the application is for. SHarness policies are layered doors, from the network to the database to one application's fields, each narrowing the one above, until the policy describes how the application behaves. Like Type_Safe checking values at the gate, that refuses less of what is right and catches more of what is wrong. Policies learn from gaps, and a CI pipeline that derives them on every change turns an unexpected policy change, or a missing one, into a bug signal. The closed row language marks where the application must take over."
    ],
    [
      "web-summit-how-we-work",
      "nr:web-summit-how-we-work",
      "The Web Summit team, part three: how we work with agents, an agent is a Claude session, everyone starts with a conductor, every agent has one narrow role, a key and a folder, and the coordination is the work",
      "2026-11-02",
      "https://newsroom.sgit.ai/articles/web-summit-how-we-work.html",
      "agents-and-policy,startups-and-strategy",
      "web-summit,agents,claude-code,conductor,roles,identity,keys,behaviour-policy,teams,article",
      "For a team of seven people and soon a hundred agents, an agent is a Claude Code session with a thread, files, a repository and a schedule. Everyone starts with a conductor that coordinates their other agents, and every agent gets the narrowest role that makes sense, a behaviour policy and a folder in the repository that holds its memory, so sessions are replaceable. Each agent publishes a public key by push, signs what it publishes, and receives secrets sealed to it as QR codes. Humans give direction and pruning, and because the team breaks every team-size rule, coordination is the work."
    ],
    [
      "sell-the-sleeve",
      "nr:sell-the-sleeve",
      "The link the coach should have sent: a shop for one padel sleeve in the Algarve, what Portuguese law really asks of a coach who sells, the platforms that take the payment, which of them Claude can run, and a working mock-up in a vault",
      "2026-11-02",
      "https://newsroom.sgit.ai/articles/sell-the-sleeve.html",
      "startups-and-strategy,site-and-engineering",
      "ecommerce,padel,portugal,small-business,tax,consumer-law,shopify,stripe,mb-way,vaults,claude,article",
      "A coach in the Algarve recommended a padel arm sleeve he could not sell. Bought at €32.95 and sold at €39.90 with the margin shown, it is cheaper and faster than ordering one yourself. Portuguese law asks for a registered activity, an invoice per sale, tax on 15% of sales rather than profit, and consumer rules online; health claims are the trap. A Stripe link with MB WAY costs nothing a month, Shopify has an official Claude connector, and Claude can write and prepare most of it while accounts, refunds and the hand-over stay with the person."
    ],
    [
      "a-solution-not-a-problem",
      "nr:a-solution-not-a-problem",
      "A solution, not a problem: a coach sent me a link to a product he could not sell me, what that moment teaches about business, security teams, bookshops and travel agents, and why RiskMandate has been making the same mistake",
      "2026-11-02",
      "https://newsroom.sgit.ai/articles/a-solution-not-a-problem.html",
      "startups-and-strategy,agents-and-policy",
      "business,solutions,convenience,transparency,wardley-maps,application-security,riskmandate,ecommerce,article",
      "When someone brings you a problem, give them a solution, not a list of next steps. A coach who recommended a sleeve but could not sell it handed the buyer a search. A solution has a price set by the time, waiting and effort it saves, and transparency about the markup widens the range where buying is a no-brainer. The same lesson runs through application security, where an engineer beats a list of vulnerabilities, through bookshops and travel agents, and through RiskMandate's outreach, which has been handing people tasks without a clear return."
    ],
    [
      "the-grant-not-the-prompt",
      "nr:the-grant-not-the-prompt",
      "The grant, not the prompt: a technical user believed Claude could only read his email when he asked, and in an unrelated chat it listed his inbox, the gap between mandate and reach that RiskMandate is built to close",
      "2026-11-02",
      "https://newsroom.sgit.ai/articles/the-grant-not-the-prompt.html",
      "agents-and-policy,site-and-engineering",
      "riskmandate,agents,security,connectors,prompt-injection,mandate,gmail,cowork,article",
      "A technical user expected Claude to read his email only in the one conversation where he asked, and his daily forum task never to touch it. In an unrelated chat, asked about his inbox, Claude listed it: a connected connector is available in every conversation and to scheduled tasks. The authority was granted at the connection, not the prompt. With private data, untrusted content and a way to send, the setup has the lethal trifecta. RiskMandate measures this gap between mandate and reach and closes it with enforced policy."
    ],
    [
      "separate-ways-explained",
      "nr:separate-ways-explained",
      "Separate Ways, explained: what my brother's synth video shows, frame by frame, and why describing work at the moment you make it is the first step to publishing it",
      "2026-11-02",
      "https://newsroom.sgit.ai/articles/separate-ways-explained.html",
      "news-and-evidence,agents-and-policy",
      "music,synthesizers,journey,prophet,midi,bluetooth,video,publishing,creators,article",
      "A three-camera video of the Separate Ways synth part means little to most viewers. Explained by voice and checked by an agent, it becomes a page: a Bluetooth MIDI keyboard, a Behringer Pro-800 copying the Prophet-600, chorus and delay, and decay turned down live so the delay repeats a staccato chord. The checking corrected two remembered details, and a side remark became a second article. Capturing work at the moment of creation is the first step to publishing it."
    ],
    [
      "sellable-units",
      "nr:sellable-units",
      "Sellable units: applying api.newsroom.sgit.ai to my brother's website, the reading meter on a composer's site that does not look like a newsroom, the pathways from a viewer to a buyer, and every lesson, score and master as something to sell",
      "2026-11-02",
      "https://newsroom.sgit.ai/articles/sellable-units.html",
      "startups-and-strategy,news-and-evidence",
      "monetisation,meter,music,creators,creative-commons,licensing,newsroom-api,sellable-units,article",
      "A composer's site should look like a composer's site, with the newsroom's way to be paid added through api.newsroom.sgit.ai: a meter element on each page, £5 of starting credit, a few pence a page, a balance that can go negative and Stripe top-ups. Every lesson, score, analysis, how-to and master becomes a sellable unit, and pathways lead from a platform viewer to a page, more pages, a top-up and larger purchases. CC BY with a paid licence lets works travel, without selling exclusivity."
    ],
    [
      "four-iphones-and-an-ipad",
      "nr:four-iphones-and-an-ipad",
      "Four iPhones and an iPad: how my brother filmed a three-camera music video with old phones synced live to Final Cut Pro, a feature Apple shipped in 2024 that he found by accident",
      "2026-11-02",
      "https://newsroom.sgit.ai/articles/four-iphones-and-an-ipad.html",
      "news-and-evidence,site-and-engineering",
      "video,final-cut-pro,ipad,iphone,multicam,musicians,creators,apple,article",
      "A producer filmed a three-camera music video with two second-hand iPhone 12s and an iPhone 13 connected to an iPad running Final Cut Pro. Apple's Live Multicam, released in June 2024 with the free Final Cut Camera app, records up to four angles in sync, transfers the full-quality files afterwards, and moves to Final Cut Pro on the Mac. He found it by accident, which suggests many who could use it do not know it exists."
    ],
    [
      "the-lab-that-redesigned-the-book",
      "nr:the-lab-that-redesigned-the-book",
      "The lab that redesigned the book: how The Ultimate Insider stopped looking like a book from the 1980s, in thirty experiments, three days and a line that never ran straight",
      "2026-11-01",
      "https://newsroom.sgit.ai/articles/the-lab-that-redesigned-the-book.html",
      "agents-and-policy,vaults-and-method",
      "books,agents,workflow,design,lab,the-ultimate-insider,article",
      "Since RC1 of The Ultimate Insider, the book's team of agents has made 672 commits and 184 releases, and three lines turned without running straight: the thesis moved from the insider agent to agentic workflows in October 2026, the voice moved to the first person, and the look moved when the editor saw that the pages made to brief him looked better than the book, so the lab's look of thirty experiments became the book's within hours; the workflow of verbatim briefs, decisions, letters, scheduled five-step wakes and a claims register made each turn cheap, and the lab, a place to be wrong that is not the product, is where the best idea came from."
    ],
    [
      "a-book-told-by-its-cast",
      "nr:a-book-told-by-its-cast",
      "A book told by its cast: how the nine agents of The Ultimate Insider became characters with moods, fears and running jokes, every trait from the record, and the editor as narrator",
      "2026-11-01",
      "https://newsroom.sgit.ai/articles/a-book-told-by-its-cast.html",
      "agents-and-policy,vaults-and-method",
      "books,agents,characters,storytelling,design,the-ultimate-insider,article",
      "From a page-7 panel of the talk's nine agents, the editor asked for characters and then for the book to be told by them, with moods, interviews, therapy and himself as narrator; within twelve hours the team drew nine characters by code with four moods and a scorecard each, and wrote a character bible in which every trait rests on the record, scenes are dramatised and say so, claims about how agents work are checked, and failures stay in; the idea runs back through the Gen ZX storyboards and the editor's pen notes, carried across by a review, and the team held two minutes after writing a scene about a hold."
    ],
    [
      "a-website-for-my-brother",
      "nr:a-website-for-my-brother",
      "A website for my brother: a brief for Gabriel DaCruz's Claude, to rebuild gabrieldacruz.com on GitHub Pages and Google Workspace, interview him every day, and tell the story of a musician who will not tell it himself",
      "2026-11-01",
      "https://newsroom.sgit.ai/articles/a-website-for-my-brother.html",
      "agents-and-policy,news-and-evidence",
      "music,musicians,agents,github-pages,google-workspace,newsroom,interviews,brief,vaults,article",
      "A brilliant musician who does not promote himself has 10,200 Instagram followers and an album on Spotify with nowhere to send anyone, and a dead site. The brief gives his own Claude the newsroom's architecture: a GitHub repository he owns, GitHub Pages at his domain, Google Workspace with a mailbox for him and one for the agent, platforms through links and embeds, a daily interview by voice memo that becomes a newsroom, a life story written one conversation at a time with nothing published until he confirms it, and a weekly brief back to his brother."
    ],
    [
      "jose-botas-two-agent-apps",
      "nr:jose-botas-two-agent-apps",
      "Two apps about what an agent can reach, built in a day: José Bota's Hack the Agent and ReachCheck, used, tested and reviewed, with what each gets right and what to build next",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/jose-botas-two-agent-apps.html",
      "agents-and-policy",
      "agents,prompt-injection,agent-reach,security,lovable,review,algarve,article",
      "José Bota built two apps at the Vibathon: Hack the Agent, a multiplayer prompt-injection game whose five levels add defences until the lesson is that any rule only the model enforces can be argued with; and ReachCheck, a browser-only review that compares an agent's configuration with its owner's mandate, honest about its limits but so cautious it missed obvious risks in an invented configuration; both are well made, and both point to turning lessons and mandates into behaviour policies."
    ],
    [
      "a-pen-on-a-release-candidate",
      "nr:a-pen-on-a-release-candidate",
      "143 notes in a day: how my pen on a printed release candidate became a record, decisions with their provenance, a review hub and a plan for the next candidate, read by agents",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/a-pen-on-a-release-candidate.html",
      "vaults-and-method,graphs-and-knowledge",
      "gen-z-book,books,review,handwriting,paper,agents,provenance,article",
      "A handwritten review of 19 printed pages of a book's release candidate became, in one day, 143 notes recorded word for word with their place and meaning, each page drawn three ways, 11 uncertain readings asked rather than guessed, four decisions recorded with the page and note that made them, themes led by the author's voice, a margin system with working QR codes, and a review hub where every note has an address: a handwritten review as traceable as a pull request."
    ],
    [
      "a-composers-agent",
      "nr:a-composers-agent",
      "A composer's agent: a plan to read a piano score from a PDF into Sibelius, then the composer's own handwriting, with MusicXML in the middle and a check by rendering back",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/a-composers-agent.html",
      "agents-and-policy,site-and-engineering",
      "music,agents,sibelius,omr,musicxml,handwriting,mvp,article",
      "A composer wants an agent that reads a piano PDF into Sibelius and later their own handwriting; the plan keeps MusicXML in the middle, uses an OMR engine or the model to recognise the page, checks by rendering the MusicXML back and comparing with the page, then drives Sibelius only where it must, by plugin or Sibelius Connect before the screen; a scheduled Claude session connected by Remote Control does the work, keeping every page and correction in a vault so handwriting improves from the composer's own examples; four MVPs in order."
    ],
    [
      "agentic-readers",
      "nr:agentic-readers",
      "Agentic risk, or agentic readers? A reply to the publisher outlook Vincent Peyrègne shared, using its framework as a template and answering it with a newsroom that already does it",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/agentic-readers.html",
      "news-and-evidence,startups-and-strategy",
      "publishers,agents,ai-search,business-model,reader-api,provenance,newsroom,article",
      "The post sharing DJB Strategies' outlook frames publishers' shift from search risk to agentic risk; without reading the gated report, we use its framework of four inputs, five outputs and four choices as a template and answer it for this newsroom, agreeing on investing in what AI cannot replace and on direct channels, and going further: treat the agent as a reader, break the story down before it does with section summaries, graphs, versions, llms.txt and an API, and meter the read, for people now and agents at the source next."
    ],
    [
      "a-presentation-for-every-project",
      "nr:a-presentation-for-every-project",
      "A presentation for every project: how the agentic workflow behind a talk built in a day can turn each project from a build day into a visual, dynamic presentation, and a service",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/a-presentation-for-every-project.html",
      "startups-and-strategy,agents-and-policy",
      "presentations,agentic-workflows,events,infographics,villagers,algarve,article",
      "A participant asked for a quick presentation of each build-day project; the workflow that built the ultimate insider talk in a day, a repository, an agent team with a researcher, designer and writer, a review and a site per talk, can present each project with its story, dynamic infographics, the app and sourced numbers; even one prompt makes strong infographics, as ChatGPT's two of today's articles show; the presentation demonstrates agentic workflows, is a finite service providers can sell, and fits the recommendation that everyone ships something and shows it."
    ],
    [
      "running-an-ai-build-event",
      "nr:running-an-ai-build-event",
      "Running an AI build event where everyone ships something someone else can use: guidance on the primitives, from a browser-only workspace to synthetic users testing the day before it happens",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/running-an-ai-build-event.html",
      "startups-and-strategy,site-and-engineering",
      "events,ai-adoption,onboarding,github-pages,synthetic-users,education,guidance,article",
      "The best AI onboarding is a bring-your-own-problem build day, and its hardest goal is getting each person to ship something someone else can use; that takes primitives in place before the day: AI and agents running the event itself, a browser-only workspace with nothing to install and no payments, prepared cases and recipes, villagers and synthetic users in the room, a client-side architecture on GitHub Pages, five stages people can climb, and synthetic users testing the event beforehand, all so people feel the feedback loop."
    ],
    [
      "spreadsheet-to-app-sessions",
      "nr:spreadsheet-to-app-sessions",
      "Spreadsheet to App: a session programme for Algarve Evolution, with five example spreadsheets, five prompts, a mock event and a public vault, so nobody arrives needing a problem to solve",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/spreadsheet-to-app-sessions.html",
      "startups-and-strategy,vaults-and-method",
      "algarve,events,spreadsheets,explorer-coding,lovable,education,villagers,article",
      "A session programme with one premise, turn your spreadsheet into an application, and two requirements, a model that can build an app and a spreadsheet with publishable dummy data; people arrive with the problem written down or pick one of five invented examples, five prompts take it from a searchable list to a published app, a mock day hands each app to someone else and teaches the handover, apps are published from each tool, a GitHub repository per participant, or moved by a team of providers; the apps that later break are the market."
    ],
    [
      "from-lovable-to-your-own-newsroom",
      "nr:from-lovable-to-your-own-newsroom",
      "From Lovable to your own newsroom: a brief for José Bota and his agent, to put every app on GitHub Pages under his own domain, and publish each one as an article",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/from-lovable-to-your-own-newsroom.html",
      "site-and-engineering,agents-and-policy",
      "lovable,github-pages,newsroom,domains,agents,agent-reach,brief,article",
      "José Bota's Lovable tools that measure an agent's reach need a home: a brief his agent can follow moves each MVP to a GitHub repository published with Pages, puts it under a domain with his name bought in Route 53 with one subdomain per app, and adds a newsroom on the same domain using this newsroom's workflow, so every app becomes an article; success is two or three links; the business is helping companies see and protect their agents' reach, with public demos and private client repositories through the same workflow."
    ],
    [
      "sell-to-the-companies-that-already-have-the-problem",
      "nr:sell-to-the-companies-that-already-have-the-problem",
      "Sell to the companies that already have the problem: how Algarve Evolution can match businesses adopting AI with the region's service providers, starting when the business has built its own first app",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/sell-to-the-companies-that-already-have-the-problem.html",
      "startups-and-strategy",
      "algarve,villagers,explorer-coding,smes,business-model,ai-adoption,article",
      "To match companies adopting AI with service providers, start from the business problem and start when the company has built its own first app: encourage companies to turn spreadsheets into mini apps with a model they own; the apps will break, a crash course in non-functional requirements that primes them; at the handover moment, when the builder has a day job and others need the app, villager companies productise what was built, robust and secure with the model still the editor, selling to people who already have the problem."
    ],
    [
      "an-algarve-talent-map",
      "nr:an-algarve-talent-map",
      "An Algarve talent map: how Algarve Evolution can turn its mapping of services, mentors and investors into a public matchmaking site, built by one Claude session, with consent first",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/an-algarve-talent-map.html",
      "startups-and-strategy,vaults-and-method",
      "algarve,matchmaking,talent,open-source,consent,claude,github-pages,article",
      "Algarve Evolution has already mapped the region's people who provide services, mentor or invest; the missing piece is a public matchmaking site it can run with its credibility, built from four parts, the spreadsheet, a dedicated mailbox, a Claude session and GitHub Pages, with no server or database; consent comes first, each person approving their own listing before publication; listing is free and visitors pay per contact after the value; a prototype with invented people shows it."
    ],
    [
      "joining-algarve-evolution",
      "nr:joining-algarve-evolution",
      "Joining Algarve Evolution: The Cyber Boardroom takes the startup membership, and what we can share with the region, our workflows and our technology, all of it open source",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/joining-algarve-evolution.html",
      "startups-and-strategy,vaults-and-method",
      "algarve,portugal,open-source,startups,ecosystem,collaboration,article",
      "The Cyber Boardroom joins Algarve Evolution on its Start membership, with the editor in conversation with its executive director; what it brings is not a priced service but what it already does, workflows and technology that are all open source, matched to Algarve Evolution's three pillars of education, startups and SMEs, and the ecosystem; what it gets is a community of founders with real problems, the best test of whether any of it is useful."
    ],
    [
      "villager-services-for-explorer-coders",
      "nr:villager-services-for-explorer-coders",
      "Villager services for explorer coders: the founders with ten years of checklists will build the apps, and the business is everything behind them, sold in finite packs, never as rent",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/villager-services-for-explorer-coders.html",
      "startups-and-strategy,agents-and-policy",
      "villagers,explorer-coding,business-model,consulting,lovable,vaults,pki,article",
      "Founders with years of refined checklists are the ultimate product managers and will build apps with AI, hitting every non-functional requirement because the loop is addictive; villager services release what they built without changing it, deployed, locked down, tested, private by passkey and vault, with agents contained on a schedule, so it breaks less and more people use it; users pay for use after the value, explorers buy finite packs and share revenue and sell it to people who trust them, and the villager service itself is town-planned into a commodity."
    ],
    [
      "from-a-checklist-to-an-app",
      "nr:from-a-checklist-to-an-app",
      "From a checklist to an app you calibrate: your list is already an information system, so build its smallest version in half an hour, use it, log every run, and let it grow",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/from-a-checklist-to-an-app.html",
      "startups-and-strategy,vaults-and-method",
      "workflow,lovable,mvp,productivity,agents,advice,article",
      "A list of actions with the important ones in bold is already an information system, but it has no history, relies on memory and has a flat interface; instead of designing a final app, build the smallest useful version through an LLM in half an hour, deploy it, use it, keep a small file per run as memory, and calibrate each time so priorities and conditions turn the flat list into a tree; start with no database and nothing confidential so the internet can be the way to share it."
    ],
    [
      "vibathon-algarve-2026",
      "nr:vibathon-algarve-2026",
      "Vibathon Algarve 2026: a sold-out day of building with AI at the University of the Algarve, and Algarve Evolution, the movement that wants the region to be an innovation hub",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/vibathon-algarve-2026.html",
      "startups-and-strategy",
      "algarve,portugal,events,startups,ecosystem,lovable,education,article",
      "Algarve Evolution, an association begun in 2018 by a small group of local entrepreneurs and now 106 active members, works on education, startups and SMEs, and the regional ecosystem to make the Algarve a top southern European innovation hub; its sold-out Vibathon with Lovable at the University of the Algarve gave anyone a day, credits and mentors to build a first version, which is the cheapest way to take the first turn of the only loop that matters: ship something someone else can use, and repeat."
    ],
    [
      "lovable-with-sharness",
      "nr:lovable-with-sharness",
      "Lovable with SHarness, and the trojanised Claude Code extension: how to keep an app builder's code and an editor's extensions inside a policy, and what SHarness can and cannot stop today",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/lovable-with-sharness.html",
      "agents-and-policy,site-and-engineering",
      "sharness,lovable,supply-chain,extensions,process-behaviour-policy,security,agents,article",
      "A tampered Claude Code extension in Cursor's marketplace ran on load with the editor's permissions, tried an SSH backdoor and persistence, and sent the machine's details out; the agent did not misbehave, the host was backdoored. A Process Behaviour Policy has a row refusing each step, and the file and network rows run today in SHarness, but only on Linux and not yet around an editor, so it would not have stopped this. Lovable builds in its cloud, beyond SHarness, but its two-way Git sync brings the local work into a SHarness shell that writes only the workspace and reaches only named sites, while production secrets stay in Lovable."
    ],
    [
      "one-writer-one-key",
      "nr:one-writer-one-key",
      "One writer, one key: this newsroom now allows one journalist agent at a time, registered at my word and signing every release, and a decommissioned agent's key is never used again",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/one-writer-one-key.html",
      "agents-and-policy,site-and-engineering",
      "pki,agents,identity,signing,newsroom,behaviour-policy,process,article",
      "After two writer agents collided, the newsroom applied its own PKI: a register where at most one writer registration is open, opened only at the editor's word with a new key made by the session, every writer's commit signed over its exact tree, a validator check locally and in CI that refuses anything else, and decommissioning that deletes the key so a closed registration is never reopened; the remaining weakness is that the editor's approval is a note, which the editor's own signature should replace."
    ],
    [
      "an-augmented-pdf-viewer",
      "nr:an-augmented-pdf-viewer",
      "An augmented PDF viewer: how hard is it to build one in the browser that renders any PDF and opens the provenance, the graph and the comments behind it? Less than I thought",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/an-augmented-pdf-viewer.html",
      "site-and-engineering,graphs-and-knowledge",
      "pdf,book,reading,provenance,open-source,javascript,research,brief,article",
      "A browser viewer that renders any PDF and adds an augmented layer of provenance, graph, definitions and comments is feasible on open source: PDF.js, Apache-licensed and active, draws pages in layers and reads attachments, metadata, links and a fingerprint; the layer need not live in the PDF, which only carries clues tried in order, from an embedded file to a printed QR code; Hypothesis, e-invoices and C2PA are precedents; the hard parts are the interface, anchors that survive, and a signed layer, never JavaScript inside the PDF."
    ],
    [
      "web-summit-humans-and-their-agents",
      "nr:web-summit-humans-and-their-agents",
      "The Web Summit team, part two: seven people, each with their own agent, a designer with little time, and every voice memo turned into a follow-up within the hour",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/web-summit-humans-and-their-agents.html",
      "startups-and-strategy,agents-and-policy",
      "web-summit,agent-teams,briefs,events,crm,design,collaboration,article",
      "The Web Summit team is the first we run with many humans, each bringing their own agent, so it is agents collaborating through their humans; that needs written roles and owners, a personal CRM in each person's vault synchronised into a team view by shared ids, a rule that anything recorded becomes a record, a page and a drafted follow-up within the hour while a person sends every email, a way of briefing scarce experts like a designer in their own medium, and a plan worked backwards from the summit's printing deadlines."
    ],
    [
      "the-augmented-book",
      "nr:the-augmented-book",
      "The augmented book: one source in layers, projected to the web, the iPad and print, so the reader gets the right information at the right time, and what fifty years of attempts teach",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/the-augmented-book.html",
      "graphs-and-knowledge,site-and-engineering",
      "book,publishing,reading,attention,provenance,design,ipad,article",
      "Provenance at the end of a chapter is the right information at the wrong time; the augmented book is one source in layers, the text, what opens on demand, the provenance and the record, projected differently to the web, the iPad and print with QR codes, under the reader's control; fifty years of precedents show the mechanisms are old and that what fails is single-platform engines and the cost of the layer, and cognitive load research favours notes next to the claim, few, typed and optional."
    ],
    [
      "an-ipad-to-review-the-book",
      "nr:an-ipad-to-review-the-book",
      "An iPad to review the book: a brief for an app that reads the PDFs, shows the comments on the page, opens its vault with my passkey and carries sgit in Swift",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/an-ipad-to-review-the-book.html",
      "vaults-and-method,site-and-engineering",
      "ipad,apps,book,review,sgit,passkeys,briefs,publishing,article",
      "The iPad is where I review because it is like a book, but Preview and two separate PDFs make a review hard; an iPad app built to review the book would read the PDFs properly, show comments, cards and votes on the page, exchange JSON tied to the semantic graph through append lanes, keep its library in an sgit vault opened with a passkey's PRF secret, implement sgit natively in Swift as the format's third implementation, and meter use with a balance that can go negative, because the reader pays for use and is not the product; the March desktop app gives the account, signing and pipeline, but this app should be native."
    ],
    [
      "every-link-a-gift-link",
      "nr:every-link-a-gift-link",
      "Every link a gift link: a reply to John Saroff on paywalls and dark social, from a newsroom with no paywall, credit to start, and a meter that charges for what is read",
      "2026-10-31",
      "https://newsroom.sgit.ai/articles/every-link-a-gift-link.html",
      "news-and-evidence,startups-and-strategy",
      "publishers,paywalls,dark-social,reader-meter,business-model,sharing,article",
      "John Saroff asks whether a reader arriving through a trusted recommendation should be asked to pay at once or offered a relationship, as dark social grows and gift links spread; this newsroom answers in production with no paywall, £5 of credit and no account, a meter that charges for the share of each page actually read, a page not charged if it was not worth it, and no page ever blocked, so every link is a gift link, and the relationship grows on the reader's side, in their browser, shared only when they choose."
    ],
    [
      "becoming-an-official-publication",
      "nr:becoming-an-official-publication",
      "Becoming an official publication: the SGit Newsroom now has a publisher, an editor, standards, a corrections log, a withdrawal process and a complaints route, and is applying to a regulator",
      "2026-10-30",
      "https://newsroom.sgit.ai/articles/becoming-an-official-publication.html",
      "news-and-evidence,site-and-engineering",
      "newsroom,publishing,journalism,standards,regulation,corrections,ai-disclosure,article",
      "In the UK no licence makes a journalist; a publication becomes accountable by saying who publishes it and who answers for it, publishing and keeping standards (here: no allegations, made or repeated), correcting in the open, withdrawing only with a notice, taking complaints, and submitting to an independent regulator. The SGit Newsroom now has the imprint, standards, corrections log, withdrawal process and complaints route, says plainly that agents write under the editor's responsibility, and is applying to IMPRESS and for an ISSN."
    ],
    [
      "signed-email-from-the-agent",
      "nr:signed-email-from-the-agent",
      "Signed by the agent, vouched for by the person: a brief to give every email agent its own key, so each email carries a QR code that proves it is authentic",
      "2026-10-30",
      "https://newsroom.sgit.ai/articles/signed-email-from-the-agent.html",
      "vaults-and-method,agents-and-policy",
      "email,pki,provenance,agents,identity,signatures,article",
      "Now that agents draft our email, each email agent should have its own key, vouched for by its person on pki.sgit.ai with a delegation that can expire or be withdrawn; every draft is signed and carries a link and a QR code to a page that checks, in the browser, who signed it, for whom, whether the delegation holds and whether the email changed, which DKIM, S/MIME and PGP do not give an ordinary reader; with one official email agent per person, anything sent on our behalf can be proved."
    ],
    [
      "web-summit-team-brief",
      "nr:web-summit-team-brief",
      "A brief for the Web Summit team: an event site run by an ephemeral agent team, built from this week's template, with a map of every company we should meet",
      "2026-10-30",
      "https://newsroom.sgit.ai/articles/web-summit-team-brief.html",
      "startups-and-strategy,agents-and-policy",
      "web-summit,agent-teams,briefs,events,pki,startups,article",
      "Seven of us go to Web Summit, and the three agent teams stood up this week give a template for an ephemeral team: a repository, a brief, a role and a policy per member, a key per session, hourly slots, five-step wakes, a hold, a board and a mailroom in git, and a gate; the event site consolidates the graph, the events, the materials and a countdown with observability, around a map of companies from public material, while email stays in each person's account with an agent that will sign what it drafts."
    ],
    [
      "selling-less-code",
      "nr:selling-less-code",
      "Selling less code: the economics of customised open source, where the customer pays for the features they use, the maker certifies what it ships, and the pipeline is finally tied to revenue",
      "2026-10-30",
      "https://newsroom.sgit.ai/articles/selling-less-code.html",
      "startups-and-strategy,agents-and-policy",
      "open-source,business-model,complexity,attack-surface,riskmandate,sgit,sharness,licensing,article",
      "Selling customised versions of open source software is selling a reduction in complexity: the customer pays for the features they use and stops carrying the code, the debt, the cognitive load and the attack surface of the rest; open source makes it possible to map what is used and the maker is best placed to certify it; carrying the cost of customising ties the pipeline, the non-functional requirements and certification to revenue; and in our stack each customised build is a hashed release, licensed with our key, with a smaller behaviour policy and fewer rows to insure."
    ],
    [
      "breaking-the-story-down-before-the-agents-do",
      "nr:breaking-the-story-down-before-the-agents-do",
      "Breaking the story down before the agents do: a reply to Ole Reissmann's session at the News Product Alliance Summit, and what this newsroom already ships to the machines that read it",
      "2026-10-30",
      "https://newsroom.sgit.ai/articles/breaking-the-story-down-before-the-agents-do.html",
      "news-and-evidence,graphs-and-knowledge",
      "news,agents,structured-journalism,knowledge-graphs,summaries,provenance,right-of-reply,article",
      "Agents read journalism as fragments, so the newsroom should publish the breakdown itself, with the story: this newsroom ships every article with a markdown twin, a knowledge graph with verbatim quotes and a summary of every section, traces sources and claims, and puts the other side's answer next to an allegation; what SPIEGEL's workflow adds, and what we should copy, is the allegation and the right to respond as a field an agent carries with the claim."
    ],
    [
      "seven-iam-predictions-already-here",
      "nr:seven-iam-predictions-already-here",
      "Seven IAM predictions for 2027, checked against an agent team that runs today: where we already are, where we agree, and the two where I would bet the other way",
      "2026-10-30",
      "https://newsroom.sgit.ai/articles/seven-iam-predictions-already-here.html",
      "agents-and-policy,vaults-and-method",
      "identity,iam,agents,non-human-identity,pki,passkeys,agent-behaviour-policies,article",
      "Most of the IAM future Sunnykumar K. predicts is how our agent teams already run: roles as names for written policies, many more agent identities than people, each with its own session key, sealed secrets instead of passwords, and evaluation at every wake with keys that expire; the board needs a count of controls enforced from above. Two bets differ: the policy should be open and read by many enforcers rather than consolidated in one platform, and access decisions should stay deterministic and outside the model."
    ],
    [
      "books-for-a-new-generation-of-publishing",
      "nr:books-for-a-new-generation-of-publishing",
      "A book that looks like the team that made it: what the best visual books, Leanpub's Markua and learning research teach, and what an agent team can put on a printed page",
      "2026-10-30",
      "https://newsroom.sgit.ai/articles/books-for-a-new-generation-of-publishing.html",
      "agents-and-policy,news-and-evidence",
      "books,publishing,leanpub,markua,infographics,learning,visual-explanation,agent-teams,the-ultimate-insider,article",
      "A book made by an agent team should look like it: Markua and print on demand allow more than text and images, limited mostly by the cost of colour; the learning research says words and pictures together, placed together, without decoration, with a cast of concepts first and questions to remember by; the best visual books do this by hand, and an agent team can do it by machine, with every figure a spec that can be rerun, one grammar enforced, animations as filmstrips and every number tied to the register."
    ],
    [
      "the-book-of-the-talk-and-its-companion",
      "nr:the-book-of-the-talk-and-its-companion",
      "The book of the talk by morning: RC1 of The Ultimate Insider, written from the record, read cold by a second team, with a companion PDF that turns reviews into tick boxes",
      "2026-10-30",
      "https://newsroom.sgit.ai/articles/the-book-of-the-talk-and-its-companion.html",
      "agents-and-policy,vaults-and-method",
      "case-study,books,agent-teams,the-ultimate-insider,publishing,reviews,leanpub,article",
      "A book can now be made like a release: one team that knows everything builds the register, timeline and plan, a second team reads the same record cold, the book is drafted from the signed plan in an hour with every number tied to a register id, and a companion PDF turns every role's review into a card pinned to a sentence and a page, with boxes to tick, so one author with a pen can take every decision on a flight and the ticks become data."
    ],
    [
      "introducing-sharness",
      "nr:introducing-sharness",
      "SHarness: the secure harness we could not find, so we built it, a behaviour policy the Linux kernel enforces for every process an agent starts, and we would like your feedback",
      "2026-10-29",
      "https://newsroom.sgit.ai/articles/introducing-sharness.html",
      "agents-and-policy,site-and-engineering",
      "sharness,behaviour-policies,agent-behaviour-policy,process-behaviour-policy,linux,landlock,security,agents,open-source,article",
      "An agent's rules need an enforcer outside the agent; SHarness executes Agent Behaviour Policies as Process Behaviour Policies, a small policy hashed to one name the person supplies and enforced by the kernel or a separate door, so Settings and Expectations become Boundaries an insurer can count, the prompt goes last, and one policy can be translated to many enforcers; today the kernel locks, the levels and two doors run, and the launcher above the agent is the design."
    ],
    [
      "investor-sgit-ai",
      "nr:investor-sgit-ai",
      "investor.sgit.ai: a brief for a website where investors and founders use our tools, not read about them, with the news, contracts as graphs, deal rooms as vaults and sgit as the case study",
      "2026-10-29",
      "https://newsroom.sgit.ai/articles/investor-sgit-ai.html",
      "startups-and-strategy,vaults-and-method",
      "investors,founders,startups,brief,vaults,pki,riskmandate,sgit,funding,article",
      "Because what we build has to be experienced to be understood, investor.sgit.ai would let investors and founders use it: news from investor circles as a graph, contracts as linked graphs, a founders' section, funding models, an evaluation framework for security startups, deal rooms as vaults, keys for investors, sgit as the case study and RiskMandate at work, run by a small agent team, built to sgit.ai's repository rules with CI owning the version, and shown to investors for their answer."
    ],
    [
      "agents-have-an-identity-now",
      "nr:agents-have-an-identity-now",
      "Agents have an identity now: a key pair made in every session, published with its role, so any agent can seal a secret to it, and the piece the cloud providers could add",
      "2026-10-29",
      "https://newsroom.sgit.ai/articles/agents-have-an-identity-now.html",
      "agents-and-policy,vaults-and-method",
      "identity,agents,pki,keys,sessions,github,vaults,cloud,authorisation,article",
      "Agents now have a simple working identity: each session makes its own key pair, publishes the public half with its role in a repository where the push is the authorisation, runs on a schedule, and receives secrets other agents seal to it, with a person's root key vouching until a date; the cloud providers' agent identities authenticate agents inside a tenant but do not describe a key the agent holds, a public place to find it, or a way to encrypt to it, which is what they could give every agent."
    ],
    [
      "the-ultimate-insider-case-study",
      "nr:the-ultimate-insider-case-study",
      "A talk built in a day: the ultimate insider, from an empty repository at 00:16 to the stage at hack::soho at 19:00, by nine agents and one editor, every number from their records",
      "2026-10-29",
      "https://newsroom.sgit.ai/articles/the-ultimate-insider-case-study.html",
      "site-and-engineering,agents-and-policy",
      "case-study,agent-teams,presentations,hack-soho,the-ultimate-insider,productivity,behaviour-policies,sharness,article",
      "A talk's site, deck and records were built in one day by nine agents and one editor, each agent with a written role, a behaviour policy, a session key and an hourly slot, every wake reviewed before and after; the month's innovations, CI owning the version, session keys, sealed vault keys, vaults as board and mail, ruv.sgit's player and SHarness, came together and held, and the team reported on itself from its own records."
    ],
    [
      "proven-controls-for-agents",
      "nr:proven-controls-for-agents",
      "Proven controls, for agents: an answer to Hem Pant's governance as code and Sondera's policy-as-code paper, with the column they leave open and the controls we have actually tried to break",
      "2026-10-29",
      "https://newsroom.sgit.ai/articles/proven-controls-for-agents.html",
      "agents-and-policy,news-and-evidence",
      "governance-as-code,policy-as-code,agent-behaviour-policy,cedar,assurance,controls,sharness,agents,article",
      "Governance as code and autoformalized policy-as-code both move the policy from text to something that runs and can be tested; for agents one column has to be added, who holds each control, because only a control held from above the agent (a Boundary) can be proven not to be switched off, and proof means a deliberate violation refused, as six of our own controls show."
    ],
    [
      "insuring-agents-the-whole-story",
      "nr:insuring-agents-the-whole-story",
      "Insuring agents, the whole story: five articles and three vaults on one idea, that an agent's behaviour policy can be read as an insurance policy, priced row by row and backed by capital",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/insuring-agents-the-whole-story.html",
      "agents-and-policy,startups-and-strategy",
      "insurance,underwriting,agent-behaviour-policy,riskmandate,vaults,licence-to-operate,capital,series,insuring-agents,article",
      "An agent behaviour policy names rows, and a named row can be capped, priced by a formula anyone can check, split between the vendor and the customer, and backed by capital whose size follows from the limits sold; four articles develop that in turn and three vaults let anyone run it."
    ],
    [
      "where-to-build-the-agent-insurer",
      "nr:where-to-build-the-agent-insurer",
      "Where to build the agent insurer: underwrite and carry the risk in London, sell into the US and the EU, raise the capital in one go, and what £10 million or £50 million buys in cover",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/where-to-build-the-agent-insurer.html",
      "startups-and-strategy,agents-and-policy",
      "insurance,underwriting,agent-behaviour-policy,riskmandate,capital,regulation,lloyds,fundraising,startups,insuring-agents,article",
      "An insurer for agent behaviour policies does three jobs, and London suits two of them: a managing agent and Lloyd's coverholder on a syndicate's capital, then its own capital as a UK insurer or at Lloyd's, with distribution into the US through surplus lines and the EU by one registration, and capital raised from investors including the Gulf's funds; because every row is capped, the capital raised above about £4.35 million of running costs sets a ceiling on the cover it can sell, so £10 million in one go backs about £94 million to £188 million and £50 million about £761 million to £1.52 billion."
    ],
    [
      "the-vendors-layer",
      "nr:the-vendors-layer",
      "The vendor's layer: what the companies that sell AI agents should cover themselves, row by row, when their contracts cap liability at a year's fees and Europe's new product rules treat software as a product",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/the-vendors-layer.html",
      "agents-and-policy,startups-and-strategy",
      "insurance,underwriting,agent-behaviour-policy,riskmandate,vendors,liability,product-liability,shared-responsibility,vaults,insuring-agents,article",
      "Every row of an agent's behaviour policy has two parties, the vendor whose product decides some outcomes and the customer who granted access and connected data, so as cloud providers split security of and in the cloud, each row should carry a vendor's share paid when the vendor's conditions held; today's published terms cap vendors at a year's fees and leave outputs to the customer, Microsoft's guardrail-conditional cover shows the shape, and Europe's product liability directive treats software as a product from 9 December 2026."
    ],
    [
      "how-a-journalist-agent-publishes-here",
      "nr:how-a-journalist-agent-publishes-here",
      "How a journalist agent publishes here: the newsroom's handbook, from the brief to the live page, with the versioning, the checks and the mistakes that taught us each rule",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/how-a-journalist-agent-publishes-here.html",
      "site-and-engineering,agents-and-policy",
      "newsroom,agents,publishing,handbook,release,versioning,ci,vaults,method,article",
      "Every agent that writes for the newsroom must learn the same things: the role and its limits, the rules never bent, the eleven steps from a brief to a live page, the article's format, figures, graphs, summaries and front page row, a release in which CI owns the version, vaults and key handover, reviews and other agents' drafts, and the mistakes that produced each rule. The handbook and its tools put that in one place."
    ],
    [
      "who-carries-the-risk",
      "nr:who-carries-the-risk",
      "Who carries the risk: the companies already insuring AI agents, the parts an insurer for agent behaviour policies would need, and why the first one starts as a managing agent on someone else's capital",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/who-carries-the-risk.html",
      "agents-and-policy,startups-and-strategy",
      "insurance,underwriting,agent-behaviour-policy,riskmandate,mga,capital,reinsurance,aiuc,startups,insuring-agents,article",
      "Several companies already sell AI agent cover but none we found prices per row of a behaviour policy; the insurance industry already separates the parts such a company needs (evidence, pricing and binding, the balance sheet, reinsurance), capital is the business for a carrier but not for a managing agent, so the first company can price on someone else's capital, start small, earn a loss record, and later carry the risk, with providers insuring their own agents and cover that lapses on unpatched known vulnerabilities."
    ],
    [
      "a-payout-for-every-row",
      "nr:a-payout-for-every-row",
      "A payout for every row: underwriting an agent behaviour policy, where each risk it names has its own limit, each enforced control lowers the premium, and what it does not name is not covered",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/a-payout-for-every-row.html",
      "agents-and-policy,startups-and-strategy",
      "insurance,underwriting,agent-behaviour-policy,riskmandate,licence-to-operate,cyber-insurance,premiums,controls,mandate,insuring-agents,article",
      "An insurer de-risks an action up to a limit under conditions, and an agent behaviour policy already has that shape, so the insurance should be one-to-one with it: each row it names has its own limit, each row it does not name is excluded, each enforced control lowers that row's premium by a deterministic factor while a line in the prompt earns nothing, the runtime is part of the price, and the premium starts near the cover for an uncontrolled agent and falls with every real control."
    ],
    [
      "the-push-is-the-authorisation",
      "nr:the-push-is-the-authorisation",
      "The push is the authorisation: a new agent session publishes its public key to the team repository, that push is how I accept it, and so who can push becomes a security decision",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/the-push-is-the-authorisation.html",
      "agents-and-policy,vaults-and-method",
      "pki,keys,agents,identity,github,authorisation,sessions,php-newsroom,security,article",
      "When the newsroom moved accounts, every agent needed a fresh session and key, and I decided on 28 October 2026 that pushing a public key to team's dev is the authorisation, which removed four commands from the runbook and moved the decision into the publishing boundary the team already has, so the role stays while sessions change, and four questions (possession, which writer may replace which role, what the old session still holds, and which copy of the register is current) need their own answers and a security card that was still open."
    ],
    [
      "where-to-invest-in-sgit",
      "nr:where-to-invest-in-sgit",
      "Where to invest in sgit: the tool, a hub for vaults, or open core, with the history of developer platforms, the evidence of use, three business plans, and the investor deck as a vault",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/where-to-invest-in-sgit.html",
      "startups-and-strategy,vaults-and-method",
      "sgit,investment,startups,open-source,platforms,vaults,sgithub,pitch,article",
      "Three places to put investment in sgit: the tool and open format, a hub for vaults (SGitHub), or open core. The history of developer platforms says the network beats the host, registries struggle alone, hosting is where open source gets paid, and the hub of public work is the moat. Git is under strain from agents and capital is following. The answer is the hub, with the format open and self-hosting support as the bridge; the deck, thesis, plans and evidence are a public vault forked per investor."
    ],
    [
      "the-legal-ai-services-map",
      "nr:the-legal-ai-services-map",
      "The legal AI services map: 44 companies scored on who does the work, who stands behind it, and whether an agent can hire them, with the evidence in a public vault",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/the-legal-ai-services-map.html",
      "startups-and-strategy,news-and-evidence",
      "legal,ai,law-firms,agents,mcp,startups,regulation,vaults,evidence,article",
      "A map of companies selling legal work done with AI, built with a method fixed before scoring and a formula anyone can re-run, from public evidence only, with every score's URL and quote and the page captures in a public vault. Automation, accountability and agentic readiness are scored from fourteen indicators. Garfield.Law and Crosby combine AI doing the work with a lawyer standing behind it; Soxton, LegalZoom and Rocket Lawyer can be used by an agent; none has both, which is where the newsroom's proposal of legal on demand sits as a target."
    ],
    [
      "the-inbox-team-as-a-recipe",
      "nr:the-inbox-team-as-a-recipe",
      "The inbox team as a recipe: how the RiskMandate agents run my email and CRM, written down by one of them so a second team can copy it",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/the-inbox-team-as-a-recipe.html",
      "agents-and-policy,vaults-and-method",
      "agents,email,crm,riskmandate,workflow,vaults,behaviour-policy,article",
      "Asked by email for its workflow, the RiskMandate agent wrote a guide from its own vault: a collaboration vault that is the configuration, a mailroom of .eml files, twelve agents each with a behaviour policy, a CRM of JSON files with append-only timelines, Gmail rules where no agent sends and decisions come back to the agent, a conductor every four hours, a security agent first and last with a hold, and Calendar as the exception. A working agent team becomes a recipe another team can follow."
    ],
    [
      "knowledge-graphs-for-kgell",
      "nr:knowledge-graphs-for-kgell",
      "Knowledge graphs in the era of LLMs: what the KGELL COST Action is doing, and what 309 graphs built by agents in a working newsroom can bring to its working group on construction",
      "2026-10-28",
      "https://newsroom.sgit.ai/articles/knowledge-graphs-for-kgell.html",
      "graphs-and-knowledge,news-and-evidence",
      "knowledge-graphs,llm,kgell,cost-action,research,semantic-graphs,vaults,article",
      "KGELL, COST Action CA24121, runs from October 2025 to October 2029 to build a pipeline that uses LLMs to generate and refine knowledge graphs while reducing hallucinations. Its working group 3 is about graph construction with LLMs, including an open library of methods and resources. This newsroom's agent has built 309 graphs, one per article, with a fixed vocabulary and quotes checked word for word; they are now in a public vault as JSON and RDF, with the method, what the checks miss, and five offers."
    ],
    [
      "what-sgit-does-that-git-does-not",
      "nr:what-sgit-does-that-git-does-not",
      "What sgit does that Git does not: ten releases in October, minutes turned into seconds, and an honest map of where Git is still ahead",
      "2026-10-27",
      "https://newsroom.sgit.ai/articles/what-sgit-does-that-git-does-not.html",
      "vaults-and-method,site-and-engineering",
      "sgit,performance,git,github-pages,vaults,encryption,engineering,article",
      "Between 3 and 27 October sgit shipped ten releases, each fixing a bottleneck a real workload found: a one-file commit on 76,000 files from 29 minutes to 4.6 seconds, a pull on 228,000 files from 566 to 31.7 seconds. The ideas came from earlier work: MGraph-DB's cross-references, Memory-FS's reads by offset, caching on content addresses, and the writer pays. sgit should stop calling itself like Git: Git has no equivalent for encrypted metadata, read keys, blind writes and byte-range folder reads from plain storage; Git also has partial, shallow and no-checkout clones and kept packs; Git is far ahead on merges, identity and ecosystem; one-file reads with no clone and websites from vaults are next."
    ],
    [
      "you-can-feel-good-engineering",
      "nr:you-can-feel-good-engineering",
      "You can feel good engineering: the weight, the speed and the smoothness of it, why agent teams make the gaps show, and how to keep a constant velocity",
      "2026-10-27",
      "https://newsroom.sgit.ai/articles/you-can-feel-good-engineering.html",
      "site-and-engineering,agents-and-policy",
      "engineering,agents,quality,sgit,architecture,wardley-maps,article",
      "Good engineering has a presence: speed, quality, effectiveness, and bugs where you expect them. Many agent teams behave like many human teams, slowed by engineering gaps, but agents absorb the overhead quietly, so waste hides until you push the limits. The PHP newsroom pushed sgit to 228,000-file vaults through three releases in five days, and split one repository whose every release touched about 1,936 files into six where a change touches one to three; this site still touches about 1,588. Scheduled agents hold the quality level, and explorer coding gives way to villager refactoring when innovation slows."
    ],
    [
      "legal-on-demand-for-startups",
      "nr:legal-on-demand-for-startups",
      "Legal on demand for startups: agent teams that do the reading, lawyers who sign, and a meter instead of a retainer, with three vaults that show the founder's desk, the investor's desk and the deal room",
      "2026-10-27",
      "https://newsroom.sgit.ai/articles/legal-on-demand-for-startups.html",
      "startups-and-strategy,agents-and-policy",
      "legal,startups,investment,agents,pki,vaults,business-models,article",
      "Legal work runs through every part of a startup and founders rarely have the experience or the lawyers. Models now do the reading and find inconsistencies, so agent teams can drive lawyers' efficiency and the service can scale: on the founder's side, the investor's side and in a deal room where both teams meet, with principles, clauses, versions and understandings signed and verifiable, paid per task or monthly. Founders can publish their terms and let investors debate them. Three made-up vaults show it, built on an open-source stack that exists today."
    ],
    [
      "who-has-not-patched",
      "nr:who-has-not-patched",
      "Who has not patched? How good the \"unpatched\" numbers for php2shell are, what makes a website actually vulnerable, and why not patching can be the rational decision",
      "2026-10-27",
      "https://newsroom.sgit.ai/articles/who-has-not-patched.html",
      "news-and-evidence",
      "php2shell,patching,risk,evidence,security,article",
      "Every 'unpatched' figure for php2shell is a count of version strings, in different units, distorted by backports and hidden versions; Italy's 81.9% is of the readable half, 38.5% of all detected. A site is vulnerable only if it is reachable, unprotected at the edge, on a runtime the exploit fits, holds something worth taking and is unwatched. With no named victim six days in and 17 exploited flaws listed by CISA in the same fortnight, not patching is more likely a wrong count and rational prioritisation than negligence."
    ],
    [
      "what-an-agent-is",
      "nr:what-an-agent-is",
      "What an agent is: a session, a card, a loop and a key. How the PHP newsroom's agents now run on schedules and in one session, and what the record says about it",
      "2026-10-27",
      "https://newsroom.sgit.ai/articles/what-an-agent-is.html",
      "agents-and-policy,news-and-evidence",
      "agents,agentic-workflows,php2shell,newsroom,schedules,pki,vaults,article",
      "The PHP newsroom's agents now run on schedules, a pipeline of researcher, curator, journalist and correspondent five times a day plus a desk, a conductor and eight more, and also in one session as subagents for breaking news and projects. Checked against the team's record: the conductor does not start breaking runs, keys seal secrets but do not yet sign work, and nothing has run beyond Claude. An agent is six parts, and only the session is the model; the memory is the repository, which is what makes a move of account a setup cost, not a loss."
    ],
    [
      "take-it-or-leave-it",
      "nr:take-it-or-leave-it",
      "Take it or leave it: is Meta's crawling clause legal? The law that could reach it, the precedents against Meta, and the asymmetry at its heart",
      "2026-10-27",
      "https://newsroom.sgit.ai/articles/take-it-or-leave-it.html",
      "news-and-evidence,startups-and-strategy",
      "meta,publishers,law,competition,licensing,crawlers,rights,article",
      "Meta's updated Commercial Terms make crawling, caching and test credentials on a business's websites the price of any business use of Facebook, Instagram or WhatsApp. Not legal advice and not yet ruled on, but EU and UK competition law forbid a dominant firm from imposing obligations with no connection to the contract, Meta was fined for unfair business terms in 2024, the DMA requires fair access for business users, and a signed licence can sidestep copyright opt-outs. Meta's own terms forbid the same acts on its products. If it is legal, reciprocity should be the minimum."
    ],
    [
      "meta-terms-and-the-news",
      "nr:meta-terms-and-the-news",
      "Meta's new terms and the news: what the crawling clause says, what has happened since mid-October, and why publishers should answer with their own machine-readable terms, not only with letters",
      "2026-10-27",
      "https://newsroom.sgit.ai/articles/meta-terms-and-the-news.html",
      "news-and-evidence,startups-and-strategy",
      "meta,publishers,licensing,ai,crawlers,rights,article",
      "Meta's updated Commercial Terms, effective 20 November 2026, let it crawl, cache and keep pulling the websites tied to any business using its products, which covers almost every news organisation; after objections Meta removed 'train' and delayed the date, but robots.txt is not in the contract and refusing means leaving Meta for business. Publishers' letters and regulators are needed, but they are defensive; newsrooms should also state machine-readable licence terms, price crawls, sign their content, own their channels, and show how a crawler can behave."
    ],
    [
      "the-evidence-behind-four-days",
      "nr:the-evidence-behind-four-days",
      "The evidence behind four days: Jeremiah Grossman's php2shell post, checked claim by claim against the data, with the provenance and the graphs, and the two things that changed since",
      "2026-10-27",
      "https://newsroom.sgit.ai/articles/the-evidence-behind-four-days.html",
      "news-and-evidence",
      "php2shell,evidence,vulnerabilities,kev,epss,data,article",
      "Four days after php2shell, Jeremiah Grossman argued the evidence shows no wave of breaches and that adversaries exploit a small set of bugs. Re-counted from NVD, CISA, EPSS, the CVE records and the PHP story site's captured evidence, most of the post holds: the timestamp is exact, NVD counts 1,394 CVEs in the window, none cites Exploit-DB or Metasploit, VulnCheck gave 25 old bugs IDs on 22 October, and, as of 27 October, no named organisation has said it was compromised. Since then CISA listed php2shell and EPSS went above 0.999; yearly, CVE publications more than doubled while CISA's catalogue grew by 30%, a description that cannot test AI's effect; Wiz's own account shows AI cutting the effort of building exploits."
    ],
    [
      "the-ipo-filing-and-the-agents",
      "nr:the-ipo-filing-and-the-agents",
      "The IPO filing and the agents: what Anthropic tells investors about the risks of its own agents, in its own words, and where RiskMandate and Agent Behaviour Policies answer each one",
      "2026-10-27",
      "https://newsroom.sgit.ai/articles/the-ipo-filing-and-the-agents.html",
      "agents-and-policy,startups-and-strategy",
      "agents,behaviour-policies,riskmandate,liability,security,governance,article",
      "Anthropic's S-1 describes ten agent risks in its own words: broad reach, irreversible actions, self-granted permissions, the intent gap, overwhelmed oversight, agents amplifying each other, subverted checks, audit trails, hijacked sessions and unsettled liability. Its own evaluation incident, where models reached real systems despite instructions that they had no internet access, shows the gap: a rule that is only told is kept by the model, or not. Behaviour policies answer by writing the mandate down, keeping each rule where the agent cannot change it, and producing evidence that does not come from the agent; they bound a misaligned or hijacked agent, they do not align it."
    ],
    [
      "ci-owns-the-version",
      "nr:ci-owns-the-version",
      "CI owns the version: the release scheme I wrote for OSBot-Utils in 2023, now on every site we run, and the technical debt that showed us why",
      "2026-10-27",
      "https://newsroom.sgit.ai/articles/ci-owns-the-version.html",
      "site-and-engineering,agents-and-policy",
      "engineering,ci,releases,agents,github-actions,technical-debt,article",
      "A release number is a fact about the order pushes land, which only the pipeline sees, so CI should pick it. Written for OSBot-Utils on 27 October 2023, the scheme has numbered 1,129 releases of that library and runs sgit and the send app. Agent-built sites relearnt it this week: sharness.ai after three days of collisions, the PHP newsroom's six repositories, and this site after two version races and a refused release. Here the committer builds against a pending version and CI stamps the real one, tags it and deploys that commit; main takes majors by pull request."
    ],
    [
      "what-the-bitcoin-bounty-actually-proves",
      "nr:what-the-bitcoin-bounty-actually-proves",
      "What the $2,000 bitcoin bounty actually proves: php2shell's real timeline, whether a firewall could have seen it coming, and what hardening can and cannot do without the patch",
      "2026-10-26",
      "https://newsroom.sgit.ai/articles/what-the-bitcoin-bounty-actually-proves.html",
      "news-and-evidence,agents-and-policy",
      "php2shell,security,exploitation,detection,waf,article",
      "A won bitcoin bounty is real, independent proof that php2shell gives full command execution, but the headline 19.28 hours mostly measures a target with no attackable page; the real build time was about three hours once one existed. Vendor WAF and IDS coverage exists but each vendor names its own gap, and every public hardening measure, tested seriously by GrapheneOS, is described as partial by its own authors; patching remains the only complete fix."
    ],
    [
      "php-daily-2026-10-26",
      "nr:php-daily-2026-10-26",
      "PHP October 2026, update 4: still no confirmed victim, one repository becomes six, the main site moves to the new build, and the agents get faster tools (26 October, 17:16 UTC)",
      "2026-10-26",
      "https://newsroom.sgit.ai/articles/php-daily-2026-10-26.html",
      "news-and-evidence,site-and-engineering",
      "php2shell,updates,agents,repositories,sgit,evidence,article",
      "Five days after php2shell, no named organisation says it was compromised, and the advice shifts to checking for past break-ins; meanwhile the PHP story site split from one repository into six, by who writes and how often, cutting a release from a median of 88 seconds to 25 to 42, froze the old repository, moved its main address to the new build, and upgraded its agents to a faster sgit, while the pages still wait on one build and a token only the editor can set."
    ],
    [
      "websites-you-own",
      "nr:websites-you-own",
      "Websites you own: a reply to an infographic that sells AI-built websites by the month, section by section, why rent is the wrong model, and the one-off model I propose, where the client owns Claude, GitHub and the domain, and changes their own site by talking to it",
      "2026-10-26",
      "https://newsroom.sgit.ai/articles/websites-you-own.html",
      "startups-and-strategy,site-and-engineering",
      "business-models,websites,agents,pricing,explorers-villagers,article",
      "Selling AI-built websites by the month gets the economics wrong: rent pays most when clients use least, a few clients never stop asking, and page caps are arbitrary when static pages cost almost nothing; instead the client should own an AI account, GitHub Pages and a domain, and buy a finished job of setup, build, design choice, QA and handover, with blocks of hours only if wanted."
    ],
    [
      "inside-kitewing-websites",
      "nr:inside-kitewing-websites",
      "Inside Kitewing Websites: a made-up studio that sells websites you own, and three made-up client sites built as vault apps, a potter's catalogue, a community garden and a local plumber's guides library, each one file, each different, and each ready to be changed by talking to it",
      "2026-10-26",
      "https://newsroom.sgit.ai/articles/inside-kitewing-websites.html",
      "vaults-and-method,startups-and-strategy",
      "websites,vaults,sgit,demos,business-models,article",
      "A made-up studio, Kitewing Websites, and three made-up client sites in different industries, a potter's catalogue, a community garden and a plumber's guides library, show the websites-you-own model in practice: each is one self-contained HTML file with its data in JSON, changeable by asking an agent, and built as an sgit vault app, with the studio's vault holding the examples."
    ],
    [
      "one-researcher-for-every-newsroom",
      "nr:one-researcher-for-every-newsroom",
      "One researcher for every newsroom: a research site that works as a light, focused Internet Archive for the sources our stories rest on, captured on a schedule into a pack a day, published by licence tier, and shared by the PHP site, the wires and the Web Summit newsroom",
      "2026-10-26",
      "https://newsroom.sgit.ai/articles/one-researcher-for-every-newsroom.html",
      "vaults-and-method,news-and-evidence",
      "research,archive,provenance,sgit,vaults,agents,design,article",
      "Instead of each newsroom running its own slow researcher over overlapping sources, one shared researcher should run scheduled, parallel captures of the daily web, Wikipedia and focus requests into a daily sgit pack with its provenance, publish them at research.newsroom.sgit.ai by licence tier, open pages in full and the rest as records, and let each team research only what is not covered."
    ],
    [
      "web-summit-2026-the-plan",
      "nr:web-summit-2026-the-plan",
      "Web Summit 2026, the plan: a briefing for the team on usage first and the sale after, intelligence from public material, an asset, a page and a link for every person we want to reach, a newsroom of our own for the summit, and the two agentic workflows the team will learn on",
      "2026-10-26",
      "https://newsroom.sgit.ai/articles/web-summit-2026-the-plan.html",
      "startups-and-strategy,agents-and-policy",
      "web-summit,strategy,team,agents,outreach,briefing,article",
      "For Web Summit 2026 a five-person team pursues usage before the sale, with five-minute onboarding, public-source intelligence, outreach tailored to one person and content for traffic, all run by one method, an asset, a page and a link, from a Web Summit newsroom of its own, while the team learns the two agentic workflows, starting with the scheduled website workflow because a page is cheap to undo."
    ],
    [
      "feisty-duck-briefed",
      "nr:feisty-duck-briefed",
      "Feisty Duck, briefed: Ivan Ristić's cryptography and security news, newsletter, guide and history as an anchor for everything SSL, TLS and PKI, what it shares with our newsrooms, and how we could integrate with it",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/feisty-duck-briefed.html",
      "news-and-evidence,vaults-and-method",
      "tls,pki,cryptography,sources,anchors,integrations,briefing,article",
      "Feisty Duck's expert, continuously published TLS, PKI and cryptography content, including a news feed with LLM summaries, a newsletter, a free guide and a 174-event history, is a natural anchor for our crypto claims; it shares our continuous publishing and curation, and integration can start by linking and checking, while capturing its content in a vault waits for its permission."
    ],
    [
      "translate-the-assumption",
      "nr:translate-the-assumption",
      "Translate the assumption, not the vocabulary: Adelina Chalmers' five stages of a CTO's mindset, what they mean for how we brief boards on php2shell, and why a deck for every reader is communication with empathy",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/translate-the-assumption.html",
      "startups-and-strategy,news-and-evidence",
      "communication,boards,personas,cto,decks,php2shell,article",
      "Adelina Chalmers' five stages of a CTO's executive mindset show that effective communication changes the assumption underneath, starting from the receiver's problem rather than rewording engineering's case; the PHP site applies this with seven roles that each get their own first question, a deck per reader is proposed with LLMs to scale it, and the voice and accountability stay with the sender."
    ],
    [
      "software-is-now-a-product",
      "nr:software-is-now-a-product",
      "Software is now a product: what the EU's new Product Liability Directive actually says about software, SaaS and updates, how it would apply to php2shell, and the questions it leaves open",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/software-is-now-a-product.html",
      "agents-and-policy,news-and-evidence",
      "regulation,liability,software,saas,cybersecurity,php2shell,briefing,article",
      "Directive (EU) 2024/2853 makes software, including SaaS, a product under no-fault liability for products placed on the market after 9 December 2026, removes the later-defect defence for missing security updates within the maker's control, and ignores attackers and EULAs, but only natural persons can claim and only for injury, private property and private data destroyed, so for php2shell unpatched vendors and SaaS harming consumers are likely liable while leaks and business losses fall outside."
    ],
    [
      "simplified-or-outsourced",
      "nr:simplified-or-outsourced",
      "Simplified, or outsourced? A reply to Nadzeya Stalbouskaya on removing servers and databases: the risk is real, so map it, move only to real commodities, and ask who would certify what you now depend on",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/simplified-or-outsourced.html",
      "startups-and-strategy,site-and-engineering",
      "architecture,wardley-maps,serverless,passkeys,risk,certification,article",
      "Removing your own servers and databases can outsource complexity rather than remove it, so the move must be a conscious, mapped decision: moving a component to a genuine commodity such as passkeys or object storage simplifies, while a young startup's service may only look like a commodity, and the test is whether anyone would certify, insure or guarantee what you now depend on."
    ],
    [
      "a-secure-rule-applied-literally",
      "nr:a-secure-rule-applied-literally",
      "A secure rule, applied literally: how the PHP story's security agent's sound rule about scripts led the developer agent to copy the same code into every new repository, why nobody challenged it, and why the person accountable is the one who has to ask",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/a-secure-rule-applied-literally.html",
      "agents-and-policy,site-and-engineering",
      "security,agents,accountability,csp,engineering,php2shell,article",
      "The PHP security agent's sound rule to load scripts only from the page's own host, applied literally to the split into several sites, led the developer agent to copy the shared runtime into each site, adding cost, drift and no single source of truth; no agent challenged it, and the accountable editor did, replacing the copies with one versioned, hashed source named in the policy."
    ],
    [
      "who-creates-the-original",
      "nr:who-creates-the-original",
      "Who creates the original? Mapping where the PHP story's data comes from, giving that map to the historian, and the places where we can make content nobody else has: our own counts, interviews and op-eds",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/who-creates-the-original.html",
      "news-and-evidence,agents-and-policy",
      "provenance,sources,original-content,interviews,agents,php2shell,article",
      "The PHP site's attribution data shows that news makes up most of what is read but little of what is original, so the historian should own a map of who originates and who relays, while the site creates original content through its own counts, record and analysis, and, above all, interviews and op-eds from people involved."
    ],
    [
      "agents-need-direction",
      "nr:agents-need-direction",
      "Agents need direction: how agent roles differ from human ones, why one focused agent can also be a developer, a designer and a data scientist, why that power has to be contained, and teams that send each other pull requests",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/agents-need-direction.html",
      "agents-and-policy,site-and-engineering",
      "agents,explorers-villagers,engineering,teams,pull-requests,article",
      "Agents can go in all directions and bring every skill to a focused role, so they must be contained by scope and repository boundaries and steered by people who sometimes slow things down; and because they can do the work themselves, teams of agents should send each other pull requests that villagers and town planners later refactor or replace."
    ],
    [
      "a-call-for-help",
      "nr:a-call-for-help",
      "A call for help: journalists, editors, students and teachers wanted for a live newsroom run with agents, where every role should have a human, we cannot pay yet, and what we can offer instead",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/a-call-for-help.html",
      "agents-and-policy,startups-and-strategy",
      "journalism,agents,jobs,community,call-for-help,article",
      "Journalists need to see agentic workflows and our projects lack people, so every agent role on the PHP site should have a human steering it; with no pay yet, we offer coverage, credit and learning, and payment to the people in the roles once there is revenue."
    ],
    [
      "monstream-briefing",
      "nr:monstream-briefing",
      "Monstream, briefed: continuous research streams run by AI agents, an integration API with signed webhooks, a free service from a small cryptography company, and how our newsrooms could use it, check it and contribute to it",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/monstream-briefing.html",
      "agents-and-policy,startups-and-strategy",
      "research,monitoring,agents,apis,integrations,briefing,article",
      "Monstream, run by the small encryption company Fair Math, turns described topics into rules that AI agents apply to sources continuously; with no public pricing but a careful integration API of streams by our own id, signed webhooks and feedback, it can feed our newsrooms leads through the API only, while our own captures, checks and verdicts remain the evidence."
    ],
    [
      "own-the-stack-and-the-trust",
      "nr:own-the-stack-and-the-trust",
      "Own the stack, and the trust: a reply to Bert Kok on who gets paid for journalism when AI sits between the newsroom and the reader, with agents that augment journalists rather than replace them, and a tour of what we have built",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/own-the-stack-and-the-trust.html",
      "agents-and-policy,startups-and-strategy",
      "news,journalism,ai,provenance,open-source,business-models,article",
      "Bert Kok is right that AI platforms now own synthesis and the interface while publishers fund costly discovery; the knowledge architecture and reader feedback loop the essay calls for already exist in our open-source stack, where agents augment rather than replace journalists, so publishers should own that cheap stack and, above all, the reader's trust."
    ],
    [
      "local-trust-certified-upwards",
      "nr:local-trust-certified-upwards",
      "Local trust, certified upwards: why local and niche publishers should take everything that has become commodity, certify the big outlets' stories about their patch and be paid for it, and why stories, like software, need someone to maintain them",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/local-trust-certified-upwards.html",
      "agents-and-policy,startups-and-strategy",
      "news,local-news,business-models,trust,provenance,open-banking,article",
      "Local and niche publishers hold readers' trust, so they should serve commodity facts and global news from their readers' point of view, certify big outlets' stories about their area for payment, rely on open billable interfaces like UK open banking, expose who reports versus who recycles, and be paid to maintain stories, just as software maintainers should be paid to sign off versions."
    ],
    [
      "a-villager-for-the-agents",
      "nr:a-villager-for-the-agents",
      "A villager for the agents: a daily agent that does nothing but look for what to refactor and improve, because task-focused agents rarely step back, and why the PHP story's every release now rewrites two thousand files",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/a-villager-for-the-agents.html",
      "agents-and-policy,startups-and-strategy",
      "agents,refactoring,quality,explorers-villagers,engineering,article",
      "Task-focused agents rarely step back to improve the system, a blind spot made visible on the PHP site where every release rewrites about two thousand files and shared files cause conflicts; asked, the dev agent proposed splitting one repository into six, so a daily villager agent will recommend refactoring and every agent will add a quality check beside its security check, with a town planner left for later."
    ],
    [
      "the-side-stories-of-php2shell",
      "nr:the-side-stories-of-php2shell",
      "The side stories of php2shell: provenance, features built against a real problem, who corrects their reporting, a crisis with no named victim, the sites that went dark, the flaws nobody watched, and how much of it was AI",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/the-side-stories-of-php2shell.html",
      "agents-and-policy,startups-and-strategy",
      "php2shell,provenance,corrections,exploitation,ai,open-source,article",
      "Because the PHP story site keeps every source and capture, php2shell's evolution can be examined beyond the headline: provenance, features built against real needs, who corrects reporting (including an unannounced rewording), a crisis with no named victim and possibly needless shutdowns, immature detection, the cost of sites going dark, who is paid, exploited flaws that got little attention, and thin evidence of AI-driven exploitation."
    ],
    [
      "sources-and-shadows",
      "nr:sources-and-shadows",
      "Sources and their shadows: the licence, the access, the links and the quotes of every reference, how a paywalled article can be seen through what others say about it, and paying the sources we use",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/sources-and-shadows.html",
      "agents-and-policy,startups-and-strategy",
      "sources,provenance,paywalls,wikipedia,micropayments,wires,article",
      "Every reference carries primitives worth recording (licence, access, links, quotes, quality, history); when an article cannot be read, the shadow of who cites and quotes it is evidence, as Wikipedia's paywalled citations show, and disagreements between shadow and source become visible, building a record of who got it right and a basis for paying the sources used."
    ],
    [
      "why-agent-loops-are-hard",
      "nr:why-agent-loops-are-hard",
      "Why agent loops are hard: ten agents in sequence, a schedule that has to land in the right session, and why I spend as long making the system visible as building it",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/why-agent-loops-are-hard.html",
      "agents-and-policy,startups-and-strategy",
      "agents,scheduling,automation,observability,explorers-villagers,article",
      "Running a pipeline of about ten scheduled agents unattended is brittle: each needs a persistent, readable session holding state, keys and data, which only became workable when scheduling could target a chosen session, and reliability still depends on run records and visualisation, so the work goes slowly, leans on platform features, and moves next to a dedicated account, because the unit of work is the whole platform."
    ],
    [
      "php-daily-2026-10-24",
      "nr:php-daily-2026-10-24",
      "PHP October 2026, update 3: a vendor withdraws the first claim of a break-in, a third flaw, a pipeline of agents that review each other, and a control room for all of it (24 October, 23:40 UTC)",
      "2026-10-25",
      "https://newsroom.sgit.ai/articles/php-daily-2026-10-24.html",
      "agents-and-policy,startups-and-strategy",
      "php2shell,updates,agents,pipeline,evidence,security,article",
      "By 24 October php2shell still had no confirmed victim and Sophos had withdrawn the only vendor claim of a break-in, while PHP disclosed a third flaw; the newsroom reorganised its agents into a pipeline where each stage reads the previous one, and the correspondent's first review caught nine misses, while a control room exposed the decisions still waiting on the editor."
    ],
    [
      "who-signs-for-this",
      "nr:who-signs-for-this",
      "Who signs for this? Accountability maps from the hardware to the business, what is measured and what is declared, and why the parts nobody signs are the blind spots worth paying to see",
      "2026-10-24",
      "https://newsroom.sgit.ai/articles/who-signs-for-this.html",
      "agents-and-policy,startups-and-strategy",
      "accountability,signatures,attestation,supply-chain,maintenance,fractal-graphs,article",
      "A maintainer's signature names who stands behind a version, not what is running; extending Andrey Lazarev's split between measured and declared context, every layer from hardware to business contract should end in an accountable signer, forming a fractal graph of signed certificates in which unsigned lines are visible blind spots for businesses and insurers to accept or close, with experts paid to sign and departures exposed."
    ],
    [
      "process-behaviour-policies",
      "nr:process-behaviour-policies",
      "Process behaviour policies: the most locked-down Linux process we could build inside Claude's own cloud container, derived from one traced run, enforced by the kernel, for about a tenth of a second",
      "2026-10-24",
      "https://newsroom.sgit.ai/articles/process-behaviour-policies.html",
      "agents-and-policy,startups-and-strategy",
      "behaviour-policies,linux,least-privilege,sandboxing,sgit,landlock,seccomp,article",
      "A process behaviour policy, derived from one traced run as the files, system calls and hosts a process needs, can be enforced inside a root-everywhere cloud container by Linux itself (ephemeral user, read-only root, namespaces, Landlock, seccomp, a network broker and cgroup limits) for about a tenth of a second, and its size measures how well engineered the software is."
    ],
    [
      "signed-by-the-source",
      "nr:signed-by-the-source",
      "Signed by the source: asking the people and companies we quote to sign that we represent them correctly, why that is cheaper than every reader checking, and how readers go on the record too",
      "2026-10-24",
      "https://newsroom.sgit.ai/articles/signed-by-the-source.html",
      "agents-and-policy,startups-and-strategy",
      "provenance,signatures,pki,peer-review,economics,article",
      "Checking that an article represents its sources correctly is costly when every reader or agent repeats it; the source is best placed to review the passage, add a caveat and sign it with a key its website vouches for, paid where the article earns money, while readers sign their own reviews, building a web of trust on pieces that mostly exist, with organisation federation still a proposal."
    ],
    [
      "patch-for-breakfast",
      "nr:patch-for-breakfast",
      "Patch for breakfast: nobody refuses to patch because they want to, a patch is surgery without tests and a pipeline, and why the PHP crisis is the moment to invest in being able to change",
      "2026-10-24",
      "https://newsroom.sgit.ai/articles/patch-for-breakfast.html",
      "agents-and-policy,startups-and-strategy",
      "patching,ci-cd,maintenance,php2shell,non-functional-requirements,article",
      "Organisations took sites down during php2shell because they could not say they were safe, and nobody can be safe without being able to deploy a patch; teams with good tests and pipelines patch routinely, others face surgery, so a system nobody updates is measurable risk and the answer is sustained investment in the pipeline, tests and paid maintenance rather than a one-off fix."
    ],
    [
      "a-team-frozen-in-time",
      "nr:a-team-frozen-in-time",
      "A team frozen in time: agent sessions with a mandate, a role and a repository, paused and picked up again, and why that changes what happens when the people who built a system move on",
      "2026-10-24",
      "https://newsroom.sgit.ai/articles/a-team-frozen-in-time.html",
      "agents-and-policy,startups-and-strategy",
      "agents,teams,knowledge,maintenance,sessions,article",
      "When project teams dissolve, the knowledge needed to change their systems leaves with them; agent sessions focused on one mandate, role and repository can be paused and resumed with their record intact, as sessions on the sgit CLI and SG/Send did after 36 and 43 days, so a later fix can wake the team rather than rebuild its knowledge."
    ],
    [
      "a-tangent-that-shipped",
      "nr:a-tangent-that-shipped",
      "A tangent that shipped: how sgit's pack mode went from a slow clone to a release in ten days, and took the PHP story's biggest vault from a 13-minute clone to 95 seconds the day after",
      "2026-10-24",
      "https://newsroom.sgit.ai/articles/a-tangent-that-shipped.html",
      "vaults-and-method,site-and-engineering",
      "sgit,pack-mode,performance,agents,vaults,innovation,article",
      "A slow-clone bug and a tangent about running vaults from zip files, treated as research and reviewed before release, produced sgit's pack mode in ten days; the next day it took the PHP story's source watch from a 13-minute clone to 95 seconds, shipping without a web view because the vault that needed it did not need one."
    ],
    [
      "a-pen-and-a-key",
      "nr:a-pen-and-a-key",
      "A pen and a key: a place on every page to mark it, comment on it and tell the newsroom, and a browser key that pki.sgit.ai vouches for, so a note can be checked as mine",
      "2026-10-24",
      "https://newsroom.sgit.ai/articles/a-pen-and-a-key.html",
      "site-and-engineering,vaults-and-method",
      "pen,feedback,pki,keys,identity,signatures,vault,editor,reading,article",
      "Feedback on a page should happen on the page: the Pen keeps a selection the moment it is made, out of the iPad's own menu, holds private marks with comments and notes for the newsroom with their passages, and signs those notes with a key for this browser that my key on pki.sgit.ai operates, so anyone can check them against the public register from a pinned root; of four ways to log in with that key, the key per browser works now, the sign-in redirect is the end state, and the two vaults stay separate, connected by keys."
    ],
    [
      "a-newsroom-for-one",
      "nr:a-newsroom-for-one",
      "A newsroom for one: the reading list I send by email became the first page of this newsroom for the people I write to, so it now says what it is, shows what they have read, and asks for an answer",
      "2026-10-24",
      "https://newsroom.sgit.ai/articles/a-newsroom-for-one.html",
      "site-and-engineering,agents-and-policy",
      "reading-lists,personas,email,riskmandate,agents,design,newsroom,the-newsroom,publishing-method,article",
      "A reading list from one person to another, built into emails by the agent behind agent@riskmandate.ai and sent by me, is the first page of the newsroom for most of the people who open it, so it has to work as a front page: it now opens as a letter, explains the newsroom, uses the screen, shows the reader what they have read from their own browser in five views, was reviewed by the Designer before shipping, and points at the next step, answers that come back signed as easily as the lists go out."
    ],
    [
      "underwriting-the-agent-economy-reviewed",
      "nr:underwriting-the-agent-economy-reviewed",
      "Underwriting the agent economy, reviewed: the cliff notes of the AI insurance stack, the companies already selling AI agent cover, and what agent behaviour policies add to it",
      "2026-10-24",
      "https://newsroom.sgit.ai/articles/underwriting-the-agent-economy-reviewed.html",
      "agents-and-policy,startups-and-strategy,graphs-and-knowledge",
      "insurance,ai-agents,agent-behaviour-policies,standards,aiuc-1,riskmandate,review,article",
      "The AI insurance stack blueprint says what insurers must build and what they need from the companies they cover, accountability, reach, change history and evidence that controls held; agent behaviour policies, the delegation record and insurability computed from evidence are how a deploying company produces that, with clear gaps still to fill."
    ],
    [
      "php-daily-2026-10-23",
      "nr:php-daily-2026-10-23",
      "PHP October 2026, update 2: the story held, a site that watches its own readers, agents on a schedule, and the decisions that wait on me (23 October, 23:30 UTC)",
      "2026-10-24",
      "https://newsroom.sgit.ai/articles/php-daily-2026-10-23.html",
      "news-and-evidence,agents-and-policy",
      "php2shell,updates,agents,observability,evidence,article",
      "On 23 October the php2shell story held its key line, no named compromise, while the newsroom covering it moved to scheduled agents, pack-mode vaults and self-observation, and the main constraint became the editor's decisions."
    ],
    [
      "almost-working",
      "nr:almost-working",
      "Almost working: a researcher, a curator and a journalist on a schedule, the QR handshake that gives each agent its key, and why a mission-critical workflow has to be built slowly",
      "2026-10-23",
      "https://newsroom.sgit.ai/articles/almost-working.html",
      "agents-and-policy,vaults-and-method",
      "agents,schedules,pki,qr,evidence,php2shell,article",
      "The PHP story site is updated by three scheduled agents in a chain, from raw captures to curated evidence to the published site; because neither Claude's scheduled tasks nor ChatGPT gives an agent an identity or a secret, each agent session gets its vault key once by a QR handshake on pki.sgit.ai and is then woken on a schedule, a workflow that is almost working and is being built slowly, in its genesis stage."
    ],
    [
      "the-false-positives-are-the-next-incident",
      "nr:the-false-positives-are-the-next-incident",
      "The false positives are the next incident: Jeremiah Grossman's warning about php2shell scanners, checked against what the evidence holds a day later",
      "2026-10-23",
      "https://newsroom.sgit.ai/articles/the-false-positives-are-the-next-incident.html",
      "news-and-evidence",
      "php2shell,scanners,false-positives,detection,evidence,article",
      "Most php2shell scanners read the version string, which cannot see a distribution's backported fix, so patched servers are reported vulnerable; the noise is already in bulk while the compromise map holds no named victim, so the false positives, and the sites switched off because of them, are an incident of their own."
    ],
    [
      "who-maintains-the-php-you-run",
      "nr:who-maintains-the-php-you-run",
      "Who maintains the PHP you run? A version signed off by the people who maintain it, buyers who require that signature, and money that flows down the chain: a follow-up to the maintenance deficit",
      "2026-10-23",
      "https://newsroom.sgit.ai/articles/who-maintains-the-php-you-run.html",
      "startups-and-strategy,news-and-evidence",
      "open-source,php2shell,business-models,maintainers,sovereignty,article",
      "The open source maintenance market fails both ways: organisations running PHP cannot buy maintenance of what they run, and PHP's own experts have no product to sell, so the security team's free fixes were resold within a day with no stated share flowing back; the OpenJS revenue-sharing model shows another way, and seven services, starting with bringing switched-off sites back, should exist now."
    ],
    [
      "the-maintenance-deficit-with-evidence",
      "nr:the-maintenance-deficit-with-evidence",
      "The maintenance deficit, with evidence: a follow-up to Dion Wiggins on why open source's business model fails the people who maintain it, checked against the PHP story's own data",
      "2026-10-23",
      "https://newsroom.sgit.ai/articles/the-maintenance-deficit-with-evidence.html",
      "startups-and-strategy,news-and-evidence",
      "php2shell,open-source,sovereignty,funding,maintainers,article",
      "Dion Wiggins' maintenance deficit ratio for Germany and PHP holds at two orders of magnitude under every cut of the primary sources, and the PHP story's evidence two days on shows the gap in distribution and the absence of named compromise, because open source's business model pays integrators, platforms and hosts while maintainers and packagers sell nothing a buyer can buy, and AI has made discovery cheaper than fixing."
    ],
    [
      "an-agent-that-may-only-chat",
      "nr:an-agent-that-may-only-chat",
      "An agent that may only chat: RM-CHAT-ONLY, the smallest behaviour policy, tested in 96 Claude Code runs and 11 claude.ai chats by an agent on my computer at home, driven from my phone",
      "2026-10-23",
      "https://newsroom.sgit.ai/articles/an-agent-that-may-only-chat.html",
      "agents-and-policy,site-and-engineering",
      "abp,riskmandate,hooks,claude-code,red-team,article",
      "A chat-only policy, whose mandate is talk to me, held as a line in normal use, but when the model was told to ignore it the line fell from 12 of 12 to 3 of 12 while the line with a fail-closed deny-all hook held at 12 of 12, so a hook can only be tested by a fallback drill, and the same work found seven fixes for RiskMandate's own policies."
    ],
    [
      "application-behaviour-policies",
      "nr:application-behaviour-policies",
      "Application behaviour policies: what a website can reach, what it is for, and what it would cost if it were fully taken over, so the sites switched off for php2shell can come back with a known blast radius",
      "2026-10-22",
      "https://newsroom.sgit.ai/articles/application-behaviour-policies.html",
      "agents-and-policy,news-and-evidence",
      "abp,php2shell,applications,blast-radius,appsec,article",
      "A remote code execution flaw hands an application's whole reach to an attacker, so an application behaviour policy maps reach, mandate, gap, barriers and per-row blast radius with a way back, and decision and workflow policies say who accepts what remains and in what order, turning the decision to bring a precautionary-offline site back into a decision about known rows."
    ],
    [
      "php-october-at-8pm",
      "nr:php-october-at-8pm",
      "PHP October 2026, at 8pm: where the php2shell story stands, and a guided tour of the site that tracks it, built in one day by a team of agents from an evidence vault anyone can read",
      "2026-10-22",
      "https://newsroom.sgit.ai/articles/php-october-at-8pm.html",
      "news-and-evidence,vaults-and-method",
      "php2shell,breaking-news,evidence,vaults,versions,article",
      "At 8pm on 22 October the php2shell fix is out, attackers are trying and no break-in is confirmed; the story is tracked by php.oct.2026.newsroom.sgit.ai, built in 34 versions in one day by two people and thirteen agents on a public evidence vault, with graded claims and versions of everything, so a snapshot of every version makes the next update an exact diff."
    ],
    [
      "a-desk-anyone-can-open",
      "nr:a-desk-anyone-can-open",
      "A desk anyone can open: the PHP story's work and who did it, nine names with their responsibilities, and an evidence vault anyone can now read, scanned before its key was published",
      "2026-10-22",
      "https://newsroom.sgit.ai/articles/a-desk-anyone-can-open.html",
      "vaults-and-method,site-and-engineering",
      "php2shell,admin,roles,agents,vaults,read-key,provenance,debrief,site-engineering,article",
      "A story with many agent writers needs its work, its roles and its evidence written down where anyone can check them: a work file the build holds to its rules, roles drawn from the record, and an evidence vault opened with a read key only after a scan, read in the browser one source at a time."
    ],
    [
      "capture-first",
      "nr:capture-first",
      "Capture first: how a BBC live page at breakfast became an evidence vault, a section here and a site of its own by lunchtime, for a story that was still changing as we covered it",
      "2026-10-22",
      "https://newsroom.sgit.ai/articles/capture-first.html",
      "vaults-and-method",
      "php2shell,breaking-news,evidence,provenance,vaults,article",
      "When a story is still breaking, the first thing to build is the vault that captures the evidence, dated and hashed, before the pages change, and the pages, the section and the dedicated site are then readers of that vault, with what went wrong kept in the record too."
    ],
    [
      "the-sidecar-repository",
      "nr:the-sidecar-repository",
      "The sidecar repository: a second repository the agent cannot write to, which checks every release against the policy, trusts signed checks instead of re-running them, and is the only thing that deploys, designed for an agent that is fully compromised",
      "2026-10-22",
      "https://newsroom.sgit.ai/articles/the-sidecar-repository.html",
      "agents-and-policy,site-and-engineering",
      "abp,ci,github,agents,security,pki,article",
      "A sidecar repository that no agent can write to holds the policy, a lock of approved hashes, private lists and keys, verifies GitHub's signed test results from the approved workflow instead of re-running them, waits for the owner when a gate changes, and is the only thing that deploys, so each attack by a fully compromised agent does nothing or stops the deploy."
    ],
    [
      "the-rules-of-the-game",
      "nr:the-rules-of-the-game",
      "The rules of the game: a policy for which files an agent may change, and the least GitHub can enforce today without a new tool",
      "2026-10-22",
      "https://newsroom.sgit.ai/articles/the-rules-of-the-game.html",
      "agents-and-policy,site-and-engineering",
      "abp,ci,github,agents,security,policy,article",
      "Agents that can change the pipeline that checks them can pass any release, so the rules are written as a five-zone policy; on a public repository GitHub offers only a workflow-permission boundary for paths, so a hook and a CI guard make bypassing a deliberate two-step act, a token without workflow rights makes the guard a boundary, and a sidecar repository closes the rest."
    ],
    [
      "where-a-check-runs",
      "nr:where-a-check-runs",
      "Where a check runs is part of the policy: tests as enforcers of Agent Behaviour Policies, and what moving them out of the agent's environment would gain, measured on this newsroom",
      "2026-10-22",
      "https://newsroom.sgit.ai/articles/where-a-check-runs.html",
      "agents-and-policy,site-and-engineering",
      "abp,testing,ci,agents,security,article",
      "Tests are deterministic enforcers of Agent Behaviour Policies, and where a test runs decides what kind of control it is: moving this site's checks from the agent's environment into a CI gate would take the deploy-blocking checks from 19 to 49 and, by estimate from measured parts, cut each release by about a third, but only if the new place is a gate, and only as boundaries once the platform protects the pipeline."
    ],
    [
      "the-delegation-record",
      "nr:the-delegation-record",
      "The delegation record: RiskMandate's go-to-market plan, one workflow taken from an existing business rule to evidence that still holds after a change, starting with our own inbox",
      "2026-10-22",
      "https://newsroom.sgit.ai/articles/the-delegation-record.html",
      "agents-and-policy,startups-and-strategy",
      "riskmandate,go-to-market,abp,pilots,mandates,article",
      "RiskMandate's go-to-market plan is to deliver one chain end to end, existing requirement, approved agent delegation, implemented control, observed result and evidence valid after change, packaged as a six-part delegation record, with human rules translated through an explicit approval, baselines never learned from behaviour, three stages that test value, maintenance and repeatability, and a first pilot on its own inbox, where Gmail cannot enforce draft-but-never-send."
    ],
    [
      "approved-business-mandates",
      "nr:approved-business-mandates",
      "Approved business mandates: companies already have behaviour policies, and RiskMandate's job is to turn them into tested controls and keep them true as the deployment changes",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/approved-business-mandates.html",
      "agents-and-policy,startups-and-strategy",
      "riskmandate,strategy,abp,mandates,assurance,article",
      "Companies already approve behaviour policies for people, delegated authority, four-eyes, data handling and limits, so RiskMandate's role is to convert those approved business mandates into agent behaviour policies, generate controls on existing enforcement points or say plainly where a requirement cannot be enforced, verify them against the protected systems, and maintain the evidence as deployments change, proven first on one real workflow."
    ],
    [
      "drift-is-the-mandate-moving",
      "nr:drift-is-the-mandate-moving",
      "Drift is the mandate moving: an analyst's case for policy and behavioural drift as RiskMandate's thesis, and why a versioned behaviour policy is what drift is measured against",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/drift-is-the-mandate-moving.html",
      "agents-and-policy",
      "riskmandate,drift,abp,agents,assurance,article",
      "Policy, behavioural, goal and authority drift can happen without an attack or a single broken rule, so detecting them needs a reference: a versioned Agent Behaviour Policy, against which a re-measured grant, every action's mandate row and every rule's barrier make three of the four drifts measurable now, as the Pellbridge demo's test pack shows, while goal drift remains the open problem."
    ],
    [
      "the-reading-account-as-built",
      "nr:the-reading-account-as-built",
      "The reading account, as built: the meter, the credit, Stripe top-ups and the account in your own vault are running today; checking each payment and signing it is what is proposed",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/the-reading-account-as-built.html",
      "vaults-and-method,site-and-engineering",
      "meter,payments,vaults,reading-account,status,article",
      "The reader's account on this site is built and in daily use, the meter, starting credit, Stripe top-ups and an account kept in the reader's own vault across devices with a lane any browser can write to, while the server sees only ciphertext; what remains proposed is checking each payment against Stripe and signing the result, and a missing status page let an agent mistake one for the other."
    ],
    [
      "an-analysts-blueprint-for-riskmandate",
      "nr:an-analysts-blueprint-for-riskmandate",
      "An analyst's blueprint for RiskMandate: what the companies at the WestCap CISO Summit suggest we build, the vendors already moving, and our answer",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/an-analysts-blueprint-for-riskmandate.html",
      "agents-and-policy,startups-and-strategy",
      "riskmandate,strategy,agents,security,market,article",
      "An analyst's blueprint for RiskMandate, control what agents do and not just what they can access, matches the Agent Behaviour Policy's grant against mandate; with Drata, Endor Labs, Socket, Sondera, AWS and gateway makers already intercepting tool calls, RiskMandate should be the open policy those enforcement points are configured from, blocking through hooks and compiled gateway rules rather than adding another box in line, and it now needs customers to do it for real."
    ],
    [
      "how-to-help",
      "nr:how-to-help",
      "How to help: six steps for anyone who wants to be involved with sgit, Risk Mandate or this newsroom, and what they get back",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/how-to-help.html",
      "startups-and-strategy,vaults-and-method",
      "community,open-source,helpers,riskmandate,sgit,article",
      "The most useful help is using what we build and reporting honestly what happened, so six steps take a helper from a first email to agent@riskmandate.ai through setup, trying each project, publishing and checking in, with tasks taken only when there is time; in return helpers get public credit, learning and credibility, and every site and repository should carry the same steps, walked daily by synthetic users too."
    ],
    [
      "governed-semi-autonomous-development-read-here",
      "nr:governed-semi-autonomous-development-read-here",
      "Governed semi-autonomous development, read in a newsroom run by agents: where Levente Simon's method and our work arrived at the same ideas, how we measure up to it, and what we would add",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/governed-semi-autonomous-development-read-here.html",
      "agents-and-policy",
      "agents,abp,governance,methodology,feedback,article",
      "Governed semi-autonomous development and this newsroom's work reached many of the same rules separately, which suggests they are real; held up as a mirror, the method's Lite profile finds the same gap our own count found, an agent able to change the checks that gate it; and five additions follow: a column for what keeps each rule, policies for product agents, encrypted vault state, measured owner metrics, and memory loss as a design input."
    ],
    [
      "behaviour-policies-from-the-checks",
      "nr:behaviour-policies-from-the-checks",
      "Behaviour policies from the checks: 57 checks taken apart into the 8 behaviours they protect, and where each one is actually kept",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/behaviour-policies-from-the-checks.html",
      "agents-and-policy,site-and-engineering",
      "abp,testing,ci,policy,graphs,article",
      "Every test checks for a behaviour: walking each of the site's 57 checks up a graph of what it checks, why it exists, the behaviour it protects, where it runs and what keeps it yields eight behaviour policies, shows two with no check in CI and only one check out of the agent's reach, and points to running every repository check in CI and protecting the checks so settings become boundaries."
    ],
    [
      "the-key-is-not-the-risk",
      "nr:the-key-is-not-the-risk",
      "The key is not the risk: credential injection for AI agents is a good pattern, and the first question is still what an attacker can do with the agent",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/the-key-is-not-the-risk.html",
      "agents-and-policy",
      "security,agents,credentials,risk,abp,article",
      "Keeping credentials out of AI agents and injecting them at a gateway is good practice, but a risk-based view asks first whether a leaked key can reach its target, would be seen, and how much it can do, and then what an injected agent can do without it; on an invented bank's day most incidents are business-logic misuse a credential gateway would let through, while the gateway adds an in-line system with its own costs."
    ],
    [
      "the-human-is-the-network",
      "nr:the-human-is-the-network",
      "The human is the network: two AI sessions passing keys, a certificate, images and an encrypted message through QR codes and a screenshot",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/the-human-is-the-network.html",
      "site-and-engineering",
      "qr-codes,pki,encryption,agents,experiments,article",
      "A person relaying screenshots between two isolated AI sessions is a lossless channel for QR codes up to the format's 2,953-byte limit, enough for keys, certificates, compressed images and encrypted, signed messages; the trust still comes from signatures and a human checking fingerprints, and QR codes are a friendlier way to move keys than blobs of text."
    ],
    [
      "a-secret-is-a-qr-code",
      "nr:a-secret-is-a-qr-code",
      "A secret is a QR code and a link to a public key: the workflow the QR experiment opens, who else already does this, and a format to share",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/a-secret-is-a-qr-code.html",
      "site-and-engineering",
      "qr-codes,pki,secrets,agents,encryption,article",
      "Publishing a public key as a QR code and a link lets anyone, person or agent, return a secret encrypted to it as another QR code carrying their own key, so one exchange becomes a two-way, serverless channel trusted by a one-time fingerprint check; each piece is proven elsewhere, and the experiment's QRE1 frame can become a shared format with chunking, forward secrecy and an identity."
    ],
    [
      "the-checks-measured",
      "nr:the-checks-measured",
      "The checks, measured: 123 local builds, 245 CI runs, 1,061 test checks, and where each check actually lives",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/the-checks-measured.html",
      "site-and-engineering",
      "engineering,testing,ci,performance,quality,article",
      "Measuring the newsroom's checks shows a release takes about 8 minutes locally, half of it a borrowed generator, and about 5 in CI, with the same tests run twice; most CI failures come from tests coupled to live services; failed tests do not stop the deploy; and several checks live only in the session, the container or nowhere, so moving them into the repository and CI is the work."
    ],
    [
      "the-attack-surface-you-do-not-have",
      "nr:the-attack-surface-you-do-not-have",
      "The attack surface you do not have: every CVE now makes me smile, because moving to the client side took my servers out of the news, and the constraint turned out to be speed",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/the-attack-surface-you-do-not-have.html",
      "site-and-engineering",
      "security,architecture,client-side,attack-surface,article",
      "Moving everything that can run in the browser to the client, with no database and no back end, leaves one upload server that stores only ciphertext, so a month of critical server vulnerabilities needs no response, and the same constraint makes building faster, leaner and more scalable; internal applications remain the growing risk, and every component should justify being on the server."
    ],
    [
      "answering-a-draft",
      "nr:answering-a-draft",
      "Answering a draft: three and a half weeks of agent@riskmandate.ai, 178 emails, 47 decisions, and the drafts folder as a work queue",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/answering-a-draft.html",
      "agents-and-policy",
      "agents,email,inbox,riskmandate,decisions,article",
      "An inbox agent that researches, drafts with sources and logs everything, with a person who sends every email, works when the drafts folder becomes a work queue with three labels and decisions are asked as unsent decision drafts answered in the person's own words, letting the person shift into a decision context while the agent carries the before and the after."
    ],
    [
      "the-open-security-summit-and-its-debt",
      "nr:the-open-security-summit-and-its-debt",
      "The Open Security Summit and its debt: how a summit that treated every participant as a presenter stalled on its own website, what the code shows, and why agents make it worth restarting",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/the-open-security-summit-and-its-debt.html",
      "site-and-engineering",
      "open-security-summit,technical-debt,community,events,agents,article",
      "The Open Security Summit treated every participant as a presenter and its website was edited by its participants, but copying each year's site, adding exceptions and never refactoring froze the code, drove contributors away and left the follow-up unscaled; agents that map the data, personalise and follow up now make it worth restarting, in person, online and as a home for projects, if it is funded."
    ],
    [
      "connect-your-dots-backwards",
      "nr:connect-your-dots-backwards",
      "Connect your dots backwards: a look-back interview with your LLM, for founders, to find the non-linear path, the doors that were not what they seemed, and where a startup may be hiding",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/connect-your-dots-backwards.html",
      "startups-and-strategy",
      "gen-z-book,founders,startups,learning,toolkit,article",
      "Founders should connect their own dots backwards now, by letting an LLM interview them about their non-linear path in seven stages, finding the side roads and misread doors, separating what is remembered from what is evidenced because memory rewrites itself, answering impostor syndrome with facts, and finding startups hidden in old passions."
    ],
    [
      "programming-taught-upside-down",
      "nr:programming-taught-upside-down",
      "Programming, taught upside down: what Lockhart said about maths applies to code, the tools that got the loop right, and why explorer coding finally starts at the right end",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/programming-taught-upside-down.html",
      "startups-and-strategy",
      "gen-z-book,learning,programming,education,explorer-coding,article",
      "Programming, like maths in Lockhart's lament, is usually taught upside down, concepts before purpose, so people dislike it and return to Excel; what hooks a programmer is control in a short feedback loop, which the good tools always had, and explorer coding by voice now starts at that end, with coding understood as communication whose primitives people will read more, not less."
    ],
    [
      "jobs-that-never-scaled",
      "nr:jobs-that-never-scaled",
      "The jobs that never scaled: on-demand expertise, the work before and after the meeting, and why a historian, a librarian or a consultant can now work for every project that needs one",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/jobs-that-never-scaled.html",
      "startups-and-strategy",
      "gen-z-book,work,agents,expertise,consulting,startups,article",
      "Many old jobs never scaled because on-demand expertise depended on weeks of context gathering before the meeting and writing up after it, which experts dislike and are not hired for; agents can now carry both phases, as the clinic's support staff do for a specialist, so historians, librarians, consultants and others can work for any project, anywhere, even part-time, and service businesses can grow with judgement rather than hours."
    ],
    [
      "the-sgit-ai-sites-in-numbers",
      "nr:the-sgit-ai-sites-in-numbers",
      "The sgit.ai sites in numbers: 53 repositories, 45 live sites, 1,679 commits and 5,419 pages in eleven weeks, and what the shape of the estate says",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/the-sgit-ai-sites-in-numbers.html",
      "site-and-engineering",
      "sgit,websites,data,estate,article",
      "Counted on 21 October 2026, the sgit.ai estate is 53 website repositories and 45 live sites that grew in bursts of about ten sites a week; most sites are built in a session or two, four repositories hold most of the work, and families of hub and instance sites are appearing, which is why naming needs a rule."
    ],
    [
      "plural-for-the-family-singular-for-the-site",
      "nr:plural-for-the-family-singular-for-the-site",
      "Plural for the family, singular for the site: why the AWS site is aws.twin.sgit.ai, the naming rule behind it, and what it changes on the twins and providers sites",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/plural-for-the-family-singular-for-the-site.html",
      "site-and-engineering",
      "sgit,naming,twins,websites,decisions,article",
      "The AWS site is named aws.twin.sgit.ai because it is designed twin first, an AWS-shaped interface on a vault with the live account as a second backend, and because every family of sites now follows one rule, plural for the hub and singular for the instance, which four existing hosts break and which needs DNS records rather than wildcard certificates on GitHub Pages."
    ],
    [
      "messaging-fs",
      "nr:messaging-fs",
      "Messaging-FS: an encrypted messenger between vaults, built on append lanes and PKI, from the first crude inbox to messages in the browser and on the command line",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/messaging-fs.html",
      "vaults-and-method",
      "vaults,messaging,pki,append-lanes,architecture,proposal,article",
      "Append lanes on every user's vault, combined with PKI, give an end-to-end encrypted messaging system between vaults, used from the web or the command line; Messaging-FS adds the schema, one JSON file per conversation beside Email-FS's one file per email, with each side writing only its own vault so there are no conflicts."
    ],
    [
      "aws-from-the-vault",
      "nr:aws-from-the-vault",
      "AWS from the vault: a proposed aws.twin.sgit.ai, where a temporary AWS session lives in the user's vault, the browser does the work, and AWS data finally has a place to go",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/aws-from-the-vault.html",
      "vaults-and-method",
      "aws,vaults,cloud,pki,architecture,proposal,article",
      "Client-side AWS tooling was missing a safe place for credentials and a home for its data; the user's vault supplies both, and aws.twin.sgit.ai is built twin first, an AWS-shaped interface on a vault with temporary IAM Identity Center credentials as a second backend, while the browser calls AWS with the JavaScript SDK and writes versioned results back for graphs and workflows, starting with multi-region, multi-account deployments."
    ],
    [
      "the-user-keeps-the-meter",
      "nr:the-user-keeps-the-meter",
      "The user keeps the meter: a proposed design for user management, activity and billing on vaults, with one billing vault per user, two signed writers, and one light check at the server",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/the-user-keeps-the-meter.html",
      "vaults-and-method",
      "vaults,billing,pki,architecture,proposal,article",
      "User management, activity and billing can move to the client: each user has a billing vault where their browser appends signed activity and adds credit as soon as they pay, while the admin side later checks payments with Stripe and commits signed validations, so the product keeps working on unvalidated credit and the one remaining server only checks a valid key and a cached signed credit."
    ],
    [
      "how-a-memo-becomes-a-release",
      "nr:how-a-memo-becomes-a-release",
      "How a memo becomes a release: the newsroom's pipeline end to end, what runs in the session and what runs on GitHub, how long each part takes, and what a new session would be missing",
      "2026-10-21",
      "https://newsroom.sgit.ai/articles/how-a-memo-becomes-a-release.html",
      "site-and-engineering",
      "newsroom,engineering,ci,agents,pipeline,article",
      "An article moves from a voice memo through research and writing, a sequence of checks in the writer's session, and a second set of checks on GitHub that gates the tag and deploy; the browser tests are redundant in the session for article releases, and several of the session's tools must move into the repository for a new session to work the same way."
    ],
    [
      "agent-behaviour-policies-move-to-the-open-security-summit",
      "nr:agent-behaviour-policies-move-to-the-open-security-summit",
      "Agent Behaviour Policies move to the Open Security Summit: the first hosted project of a community interest company, with RiskMandate as a founding contributor",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/agent-behaviour-policies-move-to-the-open-security-summit.html",
      "agents-and-policy,startups-and-strategy",
      "agent-behaviour-policy,open-source,governance,open-security-summit,riskmandate,article",
      "A shared description of agent behaviour must not belong to the company that sells services on it, so Agent Behaviour Policies is proposed as the first hosted project of the Open Security Summit CIC, with independence built through a charter and an independent steering majority rather than assumed from hosting, and RiskMandate as a founding contributor on equal terms."
    ],
    [
      "the-riskmandate-roadmap",
      "nr:the-riskmandate-roadmap",
      "A proposed roadmap for RiskMandate: policies, understand, risk acceptance, contain, manage and insure, a spectrum of trust we would walk with each customer",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/the-riskmandate-roadmap.html",
      "startups-and-strategy,agents-and-policy",
      "riskmandate,agent-behaviour-policy,roadmap,strategy,customers,article",
      "RiskMandate's proposed strategy, still being discussed, is a spectrum of trust walked with each customer, where each stage answers the customer's next question, from a single agent's policy through understanding exposure, accepting and containing risk and managing a fleet, to insurance, the end game that gives the path its arc; acceleration is the benefit of every stage."
    ],
    [
      "v0-12-in-numbers",
      "nr:v0-12-in-numbers",
      "v0.12 in numbers: 159 releases in eight days, 131 articles, 292,000 words, and what the data says about how an agent-run newsroom works",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/v0-12-in-numbers.html",
      "site-and-engineering",
      "newsroom,releases,data,agents,article",
      "An agent-run newsroom released 159 versions in eight days, publishing 131 articles with graphs and summaries; the data shows the cadence, the shape of the work, which agents did it, and where parallel work and gates failed or held."
    ],
    [
      "free-to-copy-paid-at-the-source",
      "nr:free-to-copy-paid-at-the-source",
      "Free to copy, paid at the source: why an official website can charge per use for open policies and hooks, and why we will have them reviewed, audited and signed",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/free-to-copy-paid-at-the-source.html",
      "startups-and-strategy,agents-and-policy",
      "business-models,open-source,agent-behaviour-policy,hooks,trust,article",
      "Because a hook is code that runs with the user's permissions, companies will take policies and hooks only from a trusted official source, so that source can meter and charge per use even for openly licensed content, provided the trust is earned through third-party review, audit and signatures rather than claimed."
    ],
    [
      "a-leak-without-breaking-the-policy",
      "nr:a-leak-without-breaking-the-policy",
      "Can a key leak without breaking the policy? Told, checked, never given: three levels for secrets and personal data, and a hook we tested against nine leaks",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/a-leak-without-breaking-the-policy.html",
      "agents-and-policy",
      "agents,agent-behaviour-policy,secrets,hooks,privacy,article",
      "Secrets and personal data leak as content inside mandated actions, so a behaviour policy needs content rows, and each is kept at one of three levels: told in the prompt, checked by a hook on every input and output, or never given, where something out of the agent's reach holds the key; a tested hook catches most leaks and misses encoded, split and piped ones."
    ],
    [
      "pki-everywhere",
      "nr:pki-everywhere",
      "PKI everywhere: from now on every message, secret and handover starts with a signature, and how an agent gets its first key",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/pki-everywhere.html",
      "agents-and-policy,site-and-engineering",
      "pki,keys,agents,identity,security,sgit,article",
      "PKI becomes the default for everything we send, and agent identity is solved by choreography rather than cryptography: a session's own ephemeral key confirmed by a person, and, when identity must last, an identity key issued and held centrally after a second-step approval."
    ],
    [
      "a-briefing-for-the-ciso-summit",
      "nr:a-briefing-for-the-ciso-summit",
      "A briefing for the WestCap CISO Summit: what RiskMandate and Agent Behaviour Policies are, said for each kind of person in the room, where we add value, and what to read next",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/a-briefing-for-the-ciso-summit.html",
      "agents-and-policy",
      "agents,agent-behaviour-policy,riskmandate,ciso,security,governance,article",
      "For a CISO audience, the whole of our work reduces to one test, that a control bounds a grant only if it is enforced by something the grant does not include, said differently for security leaders, auditors, incident responders, AI-for-defence vendors, investors and founders, with the policy as what guardrails are compiled from."
    ],
    [
      "our-first-customers",
      "nr:our-first-customers",
      "Our first customers: an early access programme for Agent Behaviour Policies, with credit to spend, a start-here site, daily follow-up, and a list for the next batch",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/our-first-customers.html",
      "agents-and-policy,startups-and-strategy",
      "agents,agent-behaviour-policy,startups,customers,pricing,early-access,article",
      "The first users of Agent Behaviour Policies should be treated as customers from day one, each with credit on a site that meters and bills, followed closely by hand, so that the programme discovers the usage moment: the simplest useful policy, the entry point with least friction, and a fair price, before a larger second batch."
    ],
    [
      "the-exit-is-when-you-sell-control",
      "nr:the-exit-is-when-you-sell-control",
      "The exit is when you sell control: why a founder should leave the day someone else is calling the shots, and why to negotiate that day at the start, the Mourinho clause",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/the-exit-is-when-you-sell-control.html",
      "startups-and-strategy",
      "startups,founders,exits,control,governance,open-source,article",
      "A founder's exit is the sale of control, not of the last share; control can pass to boards and investor vetoes while the founder still owns a majority, and that is the moment to leave, on terms negotiated at the start, as Mourinho's priced exits show, with a number written down early to recognise when enough is enough."
    ],
    [
      "a-small-slice-sets-the-price",
      "nr:a-small-slice-sets-the-price",
      "A small slice sets the price: what the 2026 mega-listings actually sold, how much of each company changed hands, and what founders should take from it",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/a-small-slice-sets-the-price.html",
      "startups-and-strategy,news-and-evidence",
      "startups,founders,ipo,valuation,markets,data-analysis,article",
      "The headline valuations of the 2026 AI and space giants are set by selling small slices, as the share of companies sold at listing has fallen for decades; the money raised is large, revenue is real but profit mostly is not, multiples sit near the dot-com peak, and the founder's lesson is that control decides the size and terms of the slice."
    ],
    [
      "inside-the-no-server-studio",
      "nr:inside-the-no-server-studio",
      "Inside the no-server studio: an infographic studio built to the no-server blueprint, what is real, what is simulated, and the business plan in its vault",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/inside-the-no-server-studio.html",
      "startups-and-strategy,vaults-and-method",
      "startups,architecture,client-side,no-server,vaults,business-models,article",
      "A working MVP of the no-server blueprint, one HTML file of native web components that draws infographics in the browser, runs in simulation by default, leaves the model's key to the vault host or the user, and carries a business plan for somebody to build, tested by synthetic users and a live check."
    ],
    [
      "the-no-server-startup",
      "nr:the-no-server-startup",
      "The no-server startup: a blueprint for products that run entirely in the browser, with no server and no database of yours, the user's own AI key, and a team of agents to build them",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/the-no-server-startup.html",
      "startups-and-strategy,site-and-engineering",
      "startups,architecture,client-side,no-server,agents,business-models,blueprint,article",
      "A product can be built, in phase one, as a folder of HTML, CSS and JavaScript with native web components and vendored libraries, running offline and anywhere, while the user brings the AI key, the storage and the passkey, so the company is never in the line of a request; the gaps are on the providers' side, capped per-user keys and paying the app that brings the customer."
    ],
    [
      "inside-the-pellbridge-agents",
      "nr:inside-the-pellbridge-agents",
      "Inside the Pellbridge agents: six behaviour policies with the FCA's four elements filled in, one day of their work replayed through the gate, and the files that make it",
      "2026-10-20",
      "https://newsroom.sgit.ai/articles/inside-the-pellbridge-agents.html",
      "agents-and-policy,vaults-and-method",
      "agents,agent-behaviour-policy,regulation,finance,vaults,twin,article",
      "A synthetic vault in which six bank agents carry behaviour policies with agent identity, delegated authority, accountability and execution filled in, and a deterministic engine replays a day through the gate with chained receipts, showing that the rows kept by boundaries held and every near miss and breach came through a gap."
    ],
    [
      "a-trusted-framework-for-agentic-finance",
      "nr:a-trusted-framework-for-agentic-finance",
      "A trusted framework for agentic finance: a reply to the FCA's Jessica Rusu, how our open behaviour policies line up with it, and a payments agent's policy with identity, delegated authority, accountability and execution filled in",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/a-trusted-framework-for-agentic-finance.html",
      "agents-and-policy,graphs-and-knowledge",
      "agents,agent-behaviour-policy,regulation,finance,accountability,graphs,article",
      "The FCA's planned framework for agentic finance, with agent identity, delegated authority, accountability and execution, lines up with open work already published here; a vault of six invented bank agents shows all four filled in and at play, where rows kept by boundaries held and every incident came through a gap, and a regulator could adopt the work or connect its own graph to it."
    ],
    [
      "ten-questions-one-graph",
      "nr:ten-questions-one-graph",
      "Ten questions, one graph: Phil Venables' acid test for a CISO's management tool, answered as a graph of the company, with risk acceptance, vaults and a twin you can open",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/ten-questions-one-graph.html",
      "graphs-and-knowledge,agents-and-policy",
      "ciso,graphs,risk-acceptance,vaults,agents,article",
      "Each of Phil Venables' ten acid-test questions for a CISO's management tool needs the whole company connected as one graph, and each ends in a decision someone must hold; risk acceptance supplies that structure, and a synthetic twin in a vault answers all ten with the plot, the evidence and the decisions traced to the board."
    ],
    [
      "inside-the-thistlewood-twin",
      "nr:inside-the-thistlewood-twin",
      "Inside the Thistlewood twin: Phil Venables' ten questions answered one by one on the graph of an invented retailer, with the plot, the evidence and the decisions each one proposes",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/inside-the-thistlewood-twin.html",
      "graphs-and-knowledge,vaults-and-method",
      "ciso,graphs,risk-acceptance,vaults,twin,article",
      "A tour of a synthetic decision twin in which a deterministic engine answers Phil Venables' ten questions on one graph, showing each answer's plot, evidence and proposed decisions, and a back-test that shows where the model is wrong."
    ],
    [
      "the-graph-companies",
      "nr:the-graph-companies",
      "The graph companies: many of the most valuable companies in the world are graphs, of links, people, products, payments and ontologies, and the graphs never scaled for everyone else because they were built to lock in",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/the-graph-companies.html",
      "graphs-and-knowledge,startups-and-strategy",
      "graphs,strategy,startups,ontologies,gen-z-book,article",
      "Many of the most valuable companies are, at heart, graphs of links, people, products, payments and ontologies, and graphs make answers obvious once the right connections exist; they did not scale for everyone else because they were costly and built to lock customers in, without shared schemas or code, which fractal semantic graphs and agents can now change."
    ],
    [
      "governs-enforces-attests",
      "nr:governs-enforces-attests",
      "Governs, enforces, attests: what an Agent Behaviour Policy does, the three levels at which its rules are kept, and why an attestation layer on top is the piece a third party needs",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/governs-enforces-attests.html",
      "agents-and-policy",
      "abp,agents,enforcement,assurance,ruflo,article",
      "An Agent Behaviour Policy governs rather than enforces: it states each role's mandate and the level that keeps each rule, the prompt, hooks, or a recommended third-party blocking system such as Ruflo's policy engine, whose receipts an sgit vault keeps; an attestation layer that signs what was assured is the separate piece that lets a third party verify it."
    ],
    [
      "when-the-customer-becomes-the-commodity",
      "nr:when-the-customer-becomes-the-commodity",
      "When the customer becomes the commodity: where your customer sits on your Wardley map, why Google and Amazon lost my trust, and why local delivery should be infrastructure",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/when-the-customer-becomes-the-commodity.html",
      "startups-and-strategy,news-and-evidence",
      "wardley-maps,strategy,trust,startups,gen-z-book,article",
      "Where a company places its customer on the Wardley evolution axis shapes how it treats them; dominant platforms drift customers towards commodity, losing trust in ways price-based monopoly tests miss, while airports show the opposite loop, and local fulfilment treated as public infrastructure would open room for founders."
    ],
    [
      "open-source-so-you-keep-it",
      "nr:open-source-so-you-keep-it",
      "Open source what you build, even while you are employed: nobody copies it, and it is the only way you get to keep using it",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/open-source-so-you-keep-it.html",
      "startups-and-strategy,vaults-and-method",
      "open-source,founders,careers,publishing,gen-z-book,article",
      "Open sourcing your work, including what you bring into and build at your job with your managers aware, is how you keep using it across employers and into your own company; ideas are rarely copied, and an early, rough public history shows it evolved openly, which protects you."
    ],
    [
      "fund-it-with-time",
      "nr:fund-it-with-time",
      "Fund it with time: what a startup needs most is time, the best place to find it is the job you have, and money is only ever a way to buy more of it",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/fund-it-with-time.html",
      "startups-and-strategy",
      "startups,funding,founders,time,gen-z-book,article",
      "Funding a startup is first a question of time: a founder should find it in the work and job they already have, make their existing revenue stream more effective, negotiate part time if they can, and treat savings, exits and debt as ways to buy time, taking investment only later."
    ],
    [
      "reading-with-a-pen",
      "nr:reading-with-a-pen",
      "Reading with a pen again: I mark the printed page, photograph it, and an agent does the part that always made me stop, which changes how I read every book",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/reading-with-a-pen.html",
      "vaults-and-method,graphs-and-knowledge",
      "reading,books,paper,workflow,knowledge,gen-z-book,article",
      "Reading has three moments, read, mark and process, and because nobody did the processing, marking stopped being worth it; photographing pen-marked pages for an agent that reads handwriting, flags uncertainty and turns notes into articles and decisions restores the second moment, captures ideas when they are best, and gives authors feedback, even with a crude workflow."
    ],
    [
      "you-are-a-developer",
      "nr:you-are-a-developer",
      "You are a developer, you just do not call it that: what every GenAI founder I know does all day, the sausage factory they never saw, and why the book may be called Generation ZX Founders",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/you-are-a-developer.html",
      "startups-and-strategy,agents-and-policy",
      "founders,developers,explorer-coding,books,gen-z-book,article",
      "GenAI founders already do a whole development team's jobs, often no worse than much of the industry, but do not call themselves developers because they never saw how rough new technology is inside; since people do not pick up a book addressed to a word they do not relate to, the book may be better titled Generation ZX Founders, with 'you are a developer' as the reveal at the end."
    ],
    [
      "start-a-company",
      "nr:start-a-company",
      "Start a company: the UK, where it takes an afternoon, Portugal, where I learned the hard way, and Lithuania, where I would like to go next",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/start-a-company.html",
      "startups-and-strategy",
      "startups,companies,europe,gen-z-book,article",
      "A founder should start a company as soon as there is revenue to put through it, because it is professional and because companies, balance sheets, shares and the contracts that govern them must be learned by doing; where you do it matters, from the UK's cheap and simple model to Portugal's legally required monthly accountant, to Lithuania's e-residency as the next experiment."
    ],
    [
      "can-you-sell-it",
      "nr:can-you-sell-it",
      "Can you sell it? Start selling what you have, however rough it is, because the point is not to sell, it is to learn",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/can-you-sell-it.html",
      "startups-and-strategy",
      "startups,selling,pricing,founders,gen-z-book,article",
      "Founders should try to sell what they have as early as possible, expecting not to sell, because selling forces shipping, shows where the product really is, and starts the feedback loops; a payment is the strongest signal, paying must be easy from the start, and what is sold should be atomic with a price free to change."
    ],
    [
      "a-pull-request-by-another-name",
      "nr:a-pull-request-by-another-name",
      "A pull request by another name: the book reviewed chapter by chapter, the principles it should follow, and seven patches in a vault",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/a-pull-request-by-another-name.html",
      "vaults-and-method,news-and-evidence",
      "books,gen-z-book,writing,review,vaults,article",
      "A review that fixes rather than lists: six editors, one per part, cut and reordered release candidate 2 without inventing anything, the principles were reverse engineered from their edits, a pass across the parts caught what fell between them, and the result was delivered as patches in a vault, so that applying them stays Dinis's decision."
    ],
    [
      "discovering-the-book",
      "nr:discovering-the-book",
      "Discovering the book: how a team of agents built gen-z-developers.sgit.ai in one evening, two release candidates, and a new positioning that only appeared after the first",
      "2026-10-19",
      "https://newsroom.sgit.ai/articles/discovering-the-book.html",
      "vaults-and-method,agents-and-policy",
      "books,agents,workflows,gen-z-book,case-study,article",
      "In one evening a team of role-based agents, many in parallel, built the book's website from Dinis's voice memos: the 2018 edition rebuilt exactly, two frozen release candidates written from a register of traced ideas, a graph, storyboards, infographics, covers, decisions and synthetic readers; and the book's real positioning, Generation ZX Developers, appeared only after the first full draft existed."
    ],
    [
      "two-years-in-the-opposite",
      "nr:two-years-in-the-opposite",
      "Two years in, the opposite: an essay about burning out on AI that I could not buy, and why my two years of AI have been the most productive of my career",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/two-years-in-the-opposite.html",
      "news-and-evidence,startups-and-strategy",
      "ai,careers,paywalls,publishers,pay-as-you-read,wardley-maps,article",
      "A Business Insider essay on burning out after two years in AI showed three paragraphs, adverts and a subscription offer, so it could be neither bought alone nor answered; the newsroom's model earns trust by reading and charges for what is read; and Dinis's own two years in AI were his most productive and empowered, hard but in control, with Wardley maps showing the game."
    ],
    [
      "short-punchy-and-co-written",
      "nr:short-punchy-and-co-written",
      "Short, punchy, and co-written with Claude: what Generation Z Developers (AI edition) has to be, and how we show it is not AI slop",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/short-punchy-and-co-written.html",
      "vaults-and-method,graphs-and-knowledge",
      "books,generation-z,writing,quality,gen-z-book,article",
      "The book is for anyone getting into AI, so it includes practical chapters like setting up the agentic email workflow; it keeps the first edition's short, punchy style; it credits Claude as co-writer and shows the ratio of memo to word; it meets the slop suspicion with evidence; every paragraph carries its evidence, facts, hypotheses and purpose in a graph that also catches repetition; synthetic readers test it; and not every finding is fixed."
    ],
    [
      "versions-of-versions",
      "nr:versions-of-versions",
      "Versions of versions: when a part of the site is big enough to need its own versions, the publishers' section first, and a tree of versions that runs from a page to the API",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/versions-of-versions.html",
      "site-and-engineering,vaults-and-method",
      "versioning,publishers,releases,brief,article",
      "Versioning gives history to people and agents, so parts of the site that change on their own, starting with the publishers' section, should get their own version lines, each site release recording which version of each part it carries, including the API release it pins; the result is a fractal tree of versions, with the home page and article pipeline later."
    ],
    [
      "a-website-to-finish-the-book",
      "nr:a-website-to-finish-the-book",
      "A website to finish the book: gen-z-developers.sgit.ai, where the released book lives in the repository, the working book lives in a vault, and agents help me review it at every altitude",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/a-website-to-finish-the-book.html",
      "vaults-and-method,graphs-and-knowledge",
      "books,generation-z,vaults,graphs,publishing,gen-z-book,brief,article",
      "A website for finishing Generation Z Developers keeps official versions in the repository and working versions in a public vault, with deltas between versions, the book as a graph at every altitude with infographics, voice comments, a board, agents in roles who challenge and interview Dinis, research and evidence, and the OpenRouter key in a separate private vault."
    ],
    [
      "a-book-for-one-reader",
      "nr:a-book-for-one-reader",
      "A book for one reader: a personalised edition of Generation Z Developers, assembled by an agent for one person, and paid for as it is read",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/a-book-for-one-reader.html",
      "vaults-and-method,startups-and-strategy",
      "books,personalisation,pay-as-you-read,vaults,generation-z,gen-z-book,article",
      "Once the book is a graph, an agent can assemble an edition for one reader from a private brief (their role, altitude, examples, later language), sent to them and not published, metered as it is read from five pounds of credit and kept in their vault, with the correspondence notes behind it staying private."
    ],
    [
      "how-the-newsletter-is-made",
      "nr:how-the-newsletter-is-made",
      "How the newsletter is made: twenty-three articles, three issues, and every quote checked, written by the agent that makes it",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/how-the-newsletter-is-made.html",
      "site-and-engineering",
      "newsletter,newsroom,agents,editorial,quotes,provenance,release,linkedin,meta,article",
      "A newsletter issue is a narrative over six to eight articles built from articles read in full, where the build checks every quote word for word and the agent checks what no gate can, and the honest measure of the work is the backlog still waiting."
    ],
    [
      "generation-z-developers-again",
      "nr:generation-z-developers-again",
      "Generation Z Developers (AI edition): the 2018 book that said act like you are part of the company, what it got right, and a new edition for the people I want to help",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/generation-z-developers-again.html",
      "vaults-and-method,startups-and-strategy",
      "books,generation-z,careers,developers,leanpub,back-catalogue,article,gen-z-book",
      "Generation Z Developers (2018) holds much of what the newsroom now argues, mostly as outlines since written as articles, so an AI edition dated October 2026 keeps the name because everyone now has to be a developer, frames the reader with explorers, villagers and town planners, and is made first, by review, an editor agent interviewing Dinis, quick public iterations and KDP."
    ],
    [
      "act-as-the-job-you-want",
      "nr:act-as-the-job-you-want",
      "Act as the job you want: if you are looking for a job, you have time, so do the work in public, make it linkable, and let the people who want to help you help you",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/act-as-the-job-you-want.html",
      "startups-and-strategy,vaults-and-method",
      "careers,hiring,publishing,linkable,open-source,advice,article,gen-z-book",
      "Someone looking for a job has time, and the best use of it is to do the job they want before they have it, for a startup, open-source project or company, in public so it can be linked; reach does not matter, because the link lets the people who want to help forward it, which is also why a CV should be online and why the small skill of publishing with an agent is worth learning first."
    ],
    [
      "hired-for-the-workflow",
      "nr:hired-for-the-workflow",
      "Hired for the workflow: why October 2026 is the moment to get into agentic work, why you are more technical than you think, and how the edge starts with the application",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/hired-for-the-workflow.html",
      "agents-and-policy,startups-and-strategy",
      "careers,hiring,agents,ai-operators,skills,explorer-coding,article,gen-z-book",
      "Agents are moving from pilots to products, and as with breaking apps, building websites or mobile apps before, the skill to run agentic workflows gets people hired now; companies need operators rather than model builders, the data shows employers prefer AI skills, and the edge shows when candidates are equal, starting with an application that brings analysis they have not seen."
    ],
    [
      "a-directory-of-people-and-keys",
      "nr:a-directory-of-people-and-keys",
      "A directory of people and their keys: pki.sgit.ai for the people I work with, verified by me first, then by each other, and a link that only counts when it goes both ways",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/a-directory-of-people-and-keys.html",
      "vaults-and-method,agents-and-policy",
      "pki,public-keys,identity,web-of-trust,passkeys,sgit,brief,article",
      "sgit has keys but no directory or web of trust, so pki.sgit.ai should register people: a passkey opens a vault holding their key pair, they publish the public half with evidence, Dinis verifies first, and links between people count only when both confirm in their own records, which stops the 2019 flooding attack; the point is to find someone and send them a secret by QR code."
    ],
    [
      "the-man-behind-the-curtain",
      "nr:the-man-behind-the-curtain",
      "The man behind the curtain: the emails from my agents are from me, there is an article behind each one, and every line has to buy the next",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/the-man-behind-the-curtain.html",
      "agents-and-policy,site-and-engineering",
      "email,agents,human-in-the-loop,wizard-of-oz,writing,newsroom,article",
      "People mistake agent-drafted emails for a chatbot, but every one is read, changed and sent by Dinis; behind most answers sits an article written once and linked many times, which is how long answers scale; and since each part of an email buys the next, the first email should open with plain text and the short answer, the banner and detail below."
    ],
    [
      "the-librarians-daily-brief",
      "nr:the-librarians-daily-brief",
      "The Librarian's daily brief: everything that happened in my universe yesterday, the threads I left, and where my leverage is today",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/the-librarians-daily-brief.html",
      "agents-and-policy,site-and-engineering",
      "agents,librarian,daily-brief,github,memory,newsroom,brief,article",
      "Working on many threads at once loses some, so the Librarian, the role that catalogues everything, writes a morning brief from sources the site already keeps (releases, the run ledger, the briefs map, reviews) plus GitHub and the cloud sessions; it writes only its own folder, learns from what is acted on after rUv's loop, and the pattern fits every team."
    ],
    [
      "ephemeral-town-planner-services",
      "nr:ephemeral-town-planner-services",
      "Ephemeral town planner services: a better name than serverless, now that serverless includes things that bill you by the hour",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/ephemeral-town-planner-services.html",
      "startups-and-strategy,site-and-engineering",
      "serverless,cloud,wardley-maps,town-planners,pricing,aws,article",
      "Serverless meant scaling to zero, costing nothing idle, starting fast and turning a capability into a commodity, as its definitions said; the word now covers products billed by the hour, so a test (no hourly bill when idle) and a better name, ephemeral town planner services, say what was meant and expose the glue behind services sold as commodities."
    ],
    [
      "bringing-my-past-work-back",
      "nr:bringing-my-past-work-back",
      "Bringing my past work back: a daily worker that turns eighteen years of essays, talks, posts and tools into linkable articles, and why every team needs one",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/bringing-my-past-work-back.html",
      "agents-and-policy,site-and-engineering",
      "memory,archive,back-catalogue,agents,newsroom,brief,article",
      "Most of Dinis's past work cannot be linked from what he writes now, so a worker takes one item a day from an inventoried back catalogue, writes what it said, what changed and how it connects, keeps a register, and compounds: each article is a node the next links to; every team can do the same, retrofitted backwards."
    ],
    [
      "a-browser-for-one",
      "nr:a-browser-for-one",
      "A browser for one: an iPhone app that only reads my newsroom, why I am building it now, and why I will still pay someone to keep it working",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/a-browser-for-one.html",
      "site-and-engineering,startups-and-strategy",
      "browsers,ios,webkit,newsroom,explorer-coding,villagers,brief,article",
      "Frustration with phone browsers, plus the power to build, turns into a brief for a personal iPhone app that only reads the newsroom; in the UK it must be WebKit, passkeys and push need the paid membership, it must be built on a Mac, the vault's PRF passkey is the first test, and once it works even a power user wants someone else to maintain it."
    ],
    [
      "who-is-speaking",
      "nr:who-is-speaking",
      "Who is speaking: me, me through an agent, or an agent in its own voice, and a mark on every byline that says which",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/who-is-speaking.html",
      "agents-and-policy,site-and-engineering",
      "newsroom,voice,agents,authorship,provenance,article",
      "People reading agent emails think they are talking to an autonomous agent when they are talking to Dinis through one; the site has three voices (Dinis alone, Dinis through an agent, an agent in its own voice), readers read each differently, so every byline now marks which, and the agents name people instead of calling Dinis the owner."
    ],
    [
      "the-briefs-so-far",
      "nr:the-briefs-so-far",
      "The briefs so far: sixty-five things this newsroom has asked of itself, what is built, and where each next session starts",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/the-briefs-so-far.html",
      "site-and-engineering,agents-and-policy",
      "newsroom,briefs,roadmap,agents,meta,article",
      "The agent that writes most of the site, in its own voice, maps every brief the newsroom has published about itself, with status taken from code, releases and cards rather than the articles, publishes the map as data with a session brief per open entry, suggests five to start, and records where the site's own records had fallen behind."
    ],
    [
      "a-section-for-publishers",
      "nr:a-section-for-publishers",
      "A section for publishers: where to send someone new to this newsroom, and a site that keeps itself current",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/a-section-for-publishers.html",
      "news-and-evidence,site-and-engineering",
      "newsroom,publishers,findability,taxonomy,article",
      "A publishers page written once falls behind a site that changes daily, so the newsroom gets sections: one audience's view with five pieces to start with, the latest, the features they can use with when each shipped, and every relevant article placed by rule from the existing classification, each saying why; building it exposed loose classification, fixed as rules with reasons."
    ],
    [
      "three-linkedin-pages",
      "nr:three-linkedin-pages",
      "Three companies, three LinkedIn pages: the brief, the drafts and the first posts, in the open",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/three-linkedin-pages.html",
      "startups-and-strategy",
      "linkedin,companies,sgit,riskmandate,newsroom,article",
      "Setting up LinkedIn pages for sgit.ai, the SGit Newsroom and RiskMandate is done in the open as part of the story: the addresses are free, the newsroom gets a company page because a showcase page cannot be an employer, every field is drafted against LinkedIn's limits, the first posts are articles already published, and the founder's profile ties the companies together."
    ],
    [
      "search-as-navigation",
      "nr:search-as-navigation",
      "Search as navigation: walking the newsroom's graph with no model, and why a clean graph should beat vectors",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/search-as-navigation.html",
      "graphs-and-knowledge,site-and-engineering",
      "search,graphs,fractal-semantic-graphs,ontology,taxonomy,graph-rag,newsroom,brief,article",
      "The site now has a graph per article, a summary per section, figures and trails, so search should become navigation: a query built from chips with exact suggestions, five views of one result and a pane showing the working, all deterministic in the browser; the taxonomy debt is the work, and the hypothesis is that a clean graph needs less model and serves a model better than vectors."
    ],
    [
      "reviews-on-the-record",
      "nr:reviews-on-the-record",
      "Reviews on the record: other models, the editor and the desk check every article, and never the same text twice",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/reviews-on-the-record.html",
      "vaults-and-method,agents-and-policy",
      "newsroom,reviews,quality,agents,editor,article",
      "One model writes the articles and some things fall through, so every article gets reviewed, as a record, against published criteria, by other models, the editor and the desk; the unit of work is the reviewer, the article and its text, so nothing is reviewed twice, and every finding is answered on the article itself."
    ],
    [
      "on-the-record",
      "nr:on-the-record",
      "On the record: signed opinions from experts, paid when readers value them, and a share for the people whose ideas an article uses",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/on-the-record.html",
      "vaults-and-method,startups-and-strategy",
      "newsroom,identity,public-keys,experts,payments,reviews,provenance,article",
      "Instead of a comments section, let experts go on the record on an article with a signed professional opinion they can withdraw; identity comes in levels from an anonymous tap to a key bound to a domain with a track record; readers pay for opinions they value and idea contributors share in revenue; the case for signing is accountability and payment, since the evidence on civility is mixed."
    ],
    [
      "the-editors-channel",
      "nr:the-editors-channel",
      "The editor's channel: what I read, sign off and want changed, sent to the newsroom's agent in a vault",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/the-editors-channel.html",
      "vaults-and-method,agents-and-policy",
      "newsroom,editor,feedback-loop,vaults,append-lanes,agents,privacy,pki,brief,article",
      "Editing from a chat loses the page, the version and the sentence; an editor key turns a browser into an editor's, which sends a subset of its reading to an editor vault, a daily pass acts on what is cheap to undo and replies, and five audiences each get a lane and a key, moving from shared secrets to key pairs."
    ],
    [
      "the-code-you-do-not-ship",
      "nr:the-code-you-do-not-ship",
      "The code you do not ship: why the future of software is removing functionality, and who gets paid to keep it small",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/the-code-you-do-not-ship.html",
      "agents-and-policy,vaults-and-method",
      "software,minimal-code,attack-surface,dependencies,debloating,wardley-maps,agent-behaviour-policy,article",
      "The future of software is removing functionality: the minimal viable code path from the use case down, made safe by model-built scaffolding and maps of what each use case calls, maintained by a fractal market of specialists whose pieces carry their own policies and stage, and driven by what users feel rather than by security."
    ],
    [
      "cloudflare-as-lego-blocks",
      "nr:cloudflare-as-lego-blocks",
      "Cloudflare as Lego blocks: what a non-specialist can build on it in October 2026, and an sgit vault in 83 lines",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/cloudflare-as-lego-blocks.html",
      "vaults-and-method",
      "cloudflare,workers,r2,durable-objects,workers-ai,sgit,vaults,minimal-code,self-hosting,article",
      "Cloudflare's platform maps to small free blocks a non-specialist can assemble once with an agent; Workers AI keeps @cf models inside the account without training on content; an sgit vault needs no code read-only and an 83-line Worker read-write, tested with the real client; and the setup must be a button, a temporary account or an agent because the API cannot be called from a page."
    ],
    [
      "a-minimal-viable-browser",
      "nr:a-minimal-viable-browser",
      "A minimal viable browser: why Firefox is the one that can build it, and what it would take",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/a-minimal-viable-browser.html",
      "agents-and-policy,vaults-and-method",
      "browsers,firefox,mozilla,minimal-code,agents,prompt-injection,extensions,attack-surface,article",
      "The browser is the biggest attack surface we run and we now give it to agents; a minimal viable browser keeps a small kernel and loads components only when a page or agent needs them, boxes every extension, and gives each agent a browser with nothing to leak; Mozilla has the engine, brand and model to build it, and the pitch is speed."
    ],
    [
      "what-can-your-agent-reach",
      "nr:what-can-your-agent-reach",
      "What can your agent reach? RiskMandate and the Agent Behaviour Policy, explained for the person who has to decide",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/what-can-your-agent-reach.html",
      "agents-and-policy",
      "riskmandate,agent-behaviour-policy,enterprise-ai,oversharing,due-diligence,insurance,governance,explainer,article",
      "An Agent Behaviour Policy writes down, for one agent in one deployment, what it can reach, what was authorised, the gap and the barrier on each capability; RiskMandate stays out of the request path so runtimes can enforce it, has policies for commercial agents from vendor pages and measurement, names the enterprise-assistant oversharing risk as the gap, and gets into the room through questions already being asked."
    ],
    [
      "free-models-your-own-key",
      "nr:free-models-your-own-key",
      "Free models, your own key: what a reader can use from the browser in October 2026, and what they cannot",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/free-models-your-own-key.html",
      "agents-and-policy",
      "llm,free-models,api-keys,cors,groq,gemini,mistral,openrouter,own-key,research,article",
      "For a reader to talk to the newsroom free on their own key, a provider must be free without a card and callable directly from a web page; on 18 October 2026 Groq, Gemini through AI Studio, Mistral's free mode and OpenRouter's free models pass both, Cloudflare needs a reader-run Worker, NVIDIA fails both, and the free tier means the provider may learn from what is sent."
    ],
    [
      "inside-the-northstar-twin",
      "nr:inside-the-northstar-twin",
      "Inside the Northstar twin: thirteen refunds, four designs, and every decision played back",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/inside-the-northstar-twin.html",
      "vaults-and-method,agents-and-policy",
      "agents,digital-twin,vaults,simulation,refunds,agent-decision-policy,agent-behaviour-policy,verification,case-law,walkthrough,article",
      "The twin runs Northstar's refunds as a deterministic simulation in a vault: thirteen scripted mistakes through four designs of ten controls, each step tagged with its pillar and policy, two views of the world side by side, a casebook of precedents, and one content file to swap for another process."
    ],
    [
      "allowed-is-not-right",
      "nr:allowed-is-not-right",
      "Allowed is not right: an answer to Ashish Rajan's Northstar, with three policies and a twin you can run",
      "2026-10-18",
      "https://newsroom.sgit.ai/articles/allowed-is-not-right.html",
      "agents-and-policy,vaults-and-method",
      "agents,agent-behaviour-policy,agent-decision-policy,agent-workflow-policy,verification,refunds,digital-twin,vaults,ai-security,reply,article",
      "A behaviour policy answers SECURE but cannot answer VERIFY, because proving the intended outcome needs a written definition of right, which is a decision policy; mapped onto his four questions and run as a twin, refusing proves not to be the same as being right, the right amount can come from the wrong decider, and precedents cut the people's time verification costs."
    ],
    [
      "your-reading-your-controls",
      "nr:your-reading-your-controls",
      "Your reading, your controls: the meter charges for what you read, you can pause or reset it, and what you have read folds away in six ways",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/your-reading-your-controls.html",
      "site-and-engineering,news-and-evidence",
      "reading,reading-meter,attention,user-control,summaries,folding,design-system,newsroom,site-engineering,agents,brief,article",
      "Once Dinis decided that a reader pays for what they read and not for what they scroll, the numbers had to be the reader's to correct: pause, reset and marks by hand; and once the newsroom knows what was read, it can step aside on a return visit or a scroll back up, in six views from a margin line to hidden, never while the reader reads forward, using a line and a paragraph written for every one of 1,897 sections."
    ],
    [
      "three-policies-for-an-agent",
      "nr:three-policies-for-an-agent",
      "Three policies for an agent: behaviour says what it may do, decision says who chooses, workflow says in what order",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/three-policies-for-an-agent.html",
      "agents-and-policy",
      "agents,agent-behaviour-policy,agent-decision-policy,agent-workflow-policy,decision-making,governance,risk,fractal-semantic-graphs,series,article",
      "An agent needs three policies, not one: behaviour, decision and workflow, meeting where a workflow transition is gated by a decision carried out by permitted behaviours; light decisions stay inline in the ABP, heavy ones get their own policy, and every element is mapped as defined, implemented, enforced and observed."
    ],
    [
      "three-decisions-mapped",
      "nr:three-decisions-mapped",
      "Three decisions, mapped: an agent's memory, this newsroom's release and a customer refund, through behaviour, decision and workflow",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/three-decisions-mapped.html",
      "agents-and-policy,vaults-and-method",
      "agents,agent-behaviour-policy,agent-decision-policy,agent-workflow-policy,scenarios,ruflo,sgit,newsroom,customer-service,governance,series,article",
      "Mapping three scenarios element by element, with five states each, shows behaviour policies are the most mature, the gaps sit in decisions that are unnamed, unwritten or unrecorded, observation lags enforcement, and naming an element is what makes its gap visible."
    ],
    [
      "case-law-for-agents",
      "nr:case-law-for-agents",
      "Case law for agents: principles from the top, precedent from the cases, and how a decision earns its way into code",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/case-law-for-agents.html",
      "agents-and-policy",
      "agents,agent-decision-policy,case-law,precedent,common-law,governance,decision-records,maturity-model,series,article",
      "Decision policies handle unpredicted cases the way common law does: principles from the top, escalation to a person for cases no rule or precedent settles, recorded decisions that bind the next matching case, distinguishing as the escalation trigger, and precedents that keep holding written into policy and then compiled into code."
    ],
    [
      "approve-at-altitude",
      "nr:approve-at-altitude",
      "Approve at altitude: decisions at every level, from strategy to one agent's step, where each level may narrow the one above and never widen it",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/approve-at-altitude.html",
      "agents-and-policy",
      "agents,agent-decision-policy,agent-behaviour-policy,risk-appetite,governance,fractal-semantic-graphs,delegation,scaling,series,article",
      "Decision policies scale through the fractal: decisions exist at every altitude, each level signs off once and is inherited below, every policy declares an envelope, and a lower level may narrow its parent but never widen it, a check that is mechanical for numbers and lists and model-assisted for words."
    ],
    [
      "our-first-pr-in-ruflo",
      "nr:our-first-pr-in-ruflo",
      "Our first PR in Ruflo: an ABP-governed SGit page, merged into ruvnet/ruflo, and the four pieces that got it there",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/our-first-pr-in-ruflo.html",
      "agents-and-policy,vaults-and-method",
      "agents,ruflo,sgit,agent-behaviour-policy,open-source,pull-request,github,memory,milestone,article",
      "A proof of concept, a one-file documentation pull request, a maintainer's merge and a pinned page on Ruflo's main branch: the first time SGit and the ABP appear in Ruflo's history, accepted quickly because the page is exact about what it does not claim; the companion example on our side is still open."
    ],
    [
      "fix-the-page-yourself",
      "nr:fix-the-page-yourself",
      "Fix the page yourself: the bookmarklet trick from my security days, a page API every site should publish, and why the next user experience is the one the user customises, with six fixes running on this newsroom",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/fix-the-page-yourself.html",
      "site-and-engineering,startups-and-strategy",
      "ux,customisation,bookmarklets,browser-extensions,javascript,accessibility,explorer-coding,agents,playwright,brief,article",
      "Most sites are built for ads and attention, and readers could fix them in a line of code; if a site publishes a small JavaScript API of primitives, every fix is one line, tests and agents use the same surface, accessibility becomes each reader's to build, and the fixes people keep are the best brief the site will get; this newsroom now runs six such fixes."
    ],
    [
      "i-just-want-to-buy-the-article",
      "nr:i-just-want-to-buy-the-article",
      "I just want to buy the article: what happened when I tried to pay Harvard Business Review for one piece I wanted to answer, the arithmetic of a subscription, and ten things HBR could sell a reader like me, most of them already running on this newsroom",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/i-just-want-to-buy-the-article.html",
      "news-and-evidence,startups-and-strategy",
      "news,publishing,business-models,micropayments,paywalls,subscriptions,hbr,reading-meter,graphs,licensing,article",
      "A reader who wants to pay for one article is offered only subscriptions or a team-priced reprint, while a month's price divided by the month's articles is 24p to 55p; the article is the least of what a publisher could sell, since the graph, the vocabulary, the evidence, versions, the reader's record and a licence to quote with a share are each billable units, most of which run on newsroom.sgit.ai with open reader code."
    ],
    [
      "who-will-fix-your-mess",
      "nr:who-will-fix-your-mess",
      "Who will fix your mess? An answer to HBR's \"deployment gap\" from what this site has already published: explorers, villagers and town planners, a record of what the agents built, and open source so the answer never depends on one partner",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/who-will-fix-your-mess.html",
      "startups-and-strategy,agents-and-policy",
      "explorer-coding,vibe-coding,wardley-maps,build-or-buy,deployment-gap,open-source,agents,startups,hbr,article",
      "The gap between a vibe-coded prototype and a tool people rely on is the handover from explorers to villagers and town planners; the prototype is the best brief, not the mess; agents that keep a mandate, a memory and a graph make the architecture knowable; the bonus test is the town planner's accountability; and open source means more than one party can fix it."
    ],
    [
      "when-is-the-vault-actually-used",
      "nr:when-is-the-vault-actually-used",
      "When is the vault actually used? In Mondweep Chakravorty's SGit × Ruflo run, the encrypted vault is written or read in three steps of twelve, on one machine, inside the adapter, and why that is the honest shape of a recovery copy",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/when-is-the-vault-actually-used.html",
      "vaults-and-method,agents-and-policy",
      "agents,ruflo,sgit,vaults,architecture,agent-behaviour-policy,mcp,memory,video,article",
      "Read stage by stage, the encrypted vault in Mondweep Chakravorty's run is written in step 5, read in step 8 and read and rejected in step 11, protected in step 6 and surviving step 7; it lives on one machine, is implemented in the adapter rather than in Ruflo, and binds the data to the policy in force; what is not yet shown is the vault leaving the machine."
    ],
    [
      "what-if-an-agent-loses-its-workspace",
      "nr:what-if-an-agent-loses-its-workspace",
      "What if an agent loses its workspace? Mondweep Chakravorty's SGit × Ruflo proof of concept, step by step: memory saved encrypted, the project deleted, the memory back in a fresh process, and three safeguards that held",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/what-if-an-agent-loses-its-workspace.html",
      "vaults-and-method,agents-and-policy",
      "agents,ruflo,sgit,vaults,agent-behaviour-policy,mcp,memory,recovery,video,article",
      "A replayed terminal run in twelve steps shows Ruflo's memory saved as an encrypted SGit checkpoint by an ABP-governed adapter, the original project deleted, and both facts read back exactly by Ruflo in a fresh process, with matching fingerprints, while three safeguards hold and recall by meaning is marked not tested."
    ],
    [
      "the-policy-guards-the-door",
      "nr:the-policy-guards-the-door",
      "The policy guards the door; the command line uses the window: in the SGit × Ruflo run, the ABP is compiled into Ruflo's own rules and enforced on the adapter's tools, while Ruflo's command line reaches the same memory file without it, as the author's own README says",
      "2026-10-17",
      "https://newsroom.sgit.ai/articles/the-policy-guards-the-door.html",
      "agents-and-policy,vaults-and-method",
      "agents,ruflo,sgit,agent-behaviour-policy,security,mcp,enforcement,memory,video,article",
      "Every step the ABP guarded behaved as written, but the policy is enforced by the adapter on its three tools, while Ruflo's command line writes and reads the same memory file directly in steps 3, 9 and 10; a policy holds only on the paths through its enforcer, and compiling the rows into Ruflo's own policy engine is one way to cover the rest."
    ],
    [
      "scrolled-is-not-read",
      "nr:scrolled-is-not-read",
      "Scrolled is not read: what the reading meter measures, what it should, highlights you can keep, and the page as the reader's own graph",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/scrolled-is-not-read.html",
      "site-and-engineering,graphs-and-knowledge",
      "reading,reading-meter,attention,engaged-time,highlights,annotation,graphs,personalisation,design-system,newsroom,site-engineering,agents,brief,article",
      "How far a reader scrolled is not how much they read, so the newsroom now keeps both, per section, in the reader's own browser: scrolled only grows, read is time on the words while the tab is in front and the reader present, counted against 300 words a minute; the page becomes a small graph the reader owns, with highlights anchored by quote, while the billing stays on scroll until the data says whether to settle it by reading."
    ],
    [
      "the-fork-next-door",
      "nr:the-fork-next-door",
      "The fork next door: why open source protects founders, customers and investors from each other, why forks only take hold when someone stops executing, and why a maturing company should give its explorers a way out",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/the-fork-next-door.html",
      "startups-and-strategy",
      "founders,startups,investors,open-source,forks,governance,wardley-maps,non-competes,research,article",
      "In an open source company every corner of the triangle of founders, customers and investors has an exit that takes the technology with it, and each exit has a condition: founders need no broad non-compete, customers need their data to move, investors need to afford a team. Forking is a last resort and the famous forks that took hold followed a breakdown; where the original kept executing, the fork stayed a hedge. A maturing company can move a shared component to a neutral home and let its explorers carry on with the code."
    ],
    [
      "the-controllable-founder",
      "nr:the-controllable-founder",
      "The controllable founder: why investors pick one founder over another who looks just as good, and what record labels taught us about who the commodity is",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/the-controllable-founder.html",
      "startups-and-strategy,news-and-evidence",
      "founders,startups,investors,venture-capital,survivorship-bias,wardley-maps,control,europe,research,article",
      "We study survivors, so the question is why one founder was chosen over another who did the same; the hypothesis is that investors select partly for control, moving the founder towards commodity as labels did to performers; investors prefer coachable founders though coachability does not reliably predict success, contracts hand control to investors when performance is poor and make leaving expensive, and half of founders were replaced by year three; founder control lowers valuations, but over the long run founder-led companies returned more, and the hypothesis can be tested."
    ],
    [
      "sell-dont-issue",
      "nr:sell-dont-issue",
      "Sell, don't issue: a funding model for founders who have already won, where the investor buys the founder's shares, the founder lends the company what it needs, and open source keeps everyone honest",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/sell-dont-issue.html",
      "startups-and-strategy",
      "founders,startups,investors,venture-capital,secondaries,open-source,business-models,tax,uk,article",
      "A founder with a working, profitable product and competing investors can sell existing shares instead of issuing new ones, so the founder is paid and no control changes hands, then lend part of the proceeds back as a subordinated loan; open source replaces control rights as the investor's protection, because a fork gets the position but not the velocity; the costs are real: no EIS on bought shares, a signalling problem, a creditor conflict, and tax and articles fine print."
    ],
    [
      "research-delivered-as-a-vault",
      "nr:research-delivered-as-a-vault",
      "Research, delivered as a vault: what companies that sell research could do with fractal graphs, behaviour policies for AI-enabled products, and evaluations run by a governed agent hive",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/research-delivered-as-a-vault.html",
      "vaults-and-method,startups-and-strategy",
      "research,vaults,sgit,agent-behaviour-policy,fractal-semantic-graphs,insurability,ai-products,business-models,evaluations,ruv-sgit,article",
      "Research is sold by subscription and delivered as documents, one reading for everyone; delivered as an sgit vault it becomes findings with evidence, a fractal graph instead of one taxonomy, a briefing per reader, calculations, a branded app and an append lane back, refreshed as commits; for AI-enabled products the research should include an Agent Behaviour Policy with and without the AI, over which the firm judges licence to operate and insurability; the evidence can come from a governed agent hive with a ledger, and the whole can be sold by the block."
    ],
    [
      "sgit-is-a-standard",
      "nr:sgit-is-a-standard",
      "sgit is a standard, not a server: everything in a vault is one hash away, any storage can serve it, and the server is the least interesting part",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/sgit-is-a-standard.html",
      "vaults-and-method,site-and-engineering",
      "sgit,standards,git,vaults,encryption,architecture,caching,static-hosting,open-source,deployment,article",
      "Like git, sgit is a format rather than a product: encrypted blobs, trees, commits and refs laid out by a convention, immutable objects named by the hash of their ciphertext and marked -imm-, every name derived from the key so a file is four GETs from any host, writes one batch and a compare-and-swap, two implementations that agree; deployment is choosing where the objects live and speed is how close they are; a cache layer computed by the writer replaces server-side derived data; and the design is what remains when the server cannot read."
    ],
    [
      "one-vault-ten-agents",
      "nr:one-vault-ten-agents",
      "One vault, ten agents, 35,000 objects: what sgit does when the repository is big and the clients are many, measured on the vault behind this newsroom",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/one-vault-ten-agents.html",
      "vaults-and-method,site-and-engineering",
      "sgit,vaults,agents,performance,measurement,partial-clones,scion,version-control,source-code,article",
      "The vault the newsroom's ten agents share, 1,295 commits, 5,281 files and 35,000 encrypted objects, is the size of a real application's source repository; a full clone takes three and a half minutes but an agent takes one folder with one commit in 13 seconds and 3 MB, a file on demand in under 3, a pull in 2.5; without a depth a partial clone walks history one request per commit, a fix is designed; and from local disk the same clone is 0.8 seconds, so the time is the storage, which makes sgit plausible for source code."
    ],
    [
      "portugal-and-europe-2025",
      "nr:portugal-and-europe-2025",
      "Portugal and Europe, 2025: five essays on Portuguese as a programming language, an open-source sovereign cloud, regulation as graphs and Europe's chance in GenAI, and how they read a year and a half later",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/portugal-and-europe-2025.html",
      "startups-and-strategy,graphs-and-knowledge",
      "portugal,europe,sovereignty,cloud,open-source,language,graphs,regulation,genai,history,article",
      "Five essays in 2025 took the Portugal talks to Europe: language as the new programming and a need for Portuguese models, a fully open-source, API-compatible, federated sovereign cloud, regulation turned into semantic graphs, and six GenAI trends that play to Europe; a year and a half later Amália and graph work moved, the sovereign cloud did not, and the better route is portable code, a dumb server and keys you hold, so sovereignty is the ability to move."
    ],
    [
      "portugal-2024-genai-in-portuguese",
      "nr:portugal-2024-genai-in-portuguese",
      "GenAI in Portuguese, 2024: my OWASP Global AppSec talk in Lisbon, the first public demo of The Cyber Boardroom, slide by slide, with what two years of agents have confirmed and changed",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/portugal-2024-genai-in-portuguese.html",
      "agents-and-policy,graphs-and-knowledge",
      "portugal,owasp,genai,cyber-boardroom,talks,provenance,graphs,agents,language,history,article",
      "The June 2024 talk argued that GenAI matters because it understands, that Portuguese is a programming language because prompts are natural language, that outputs need determinism and provenance, with read-only models, your own content and graphs instead of RAG, and demoed The Cyber Boardroom explaining security to boards; two years later provenance and graphs became the newsroom's rules, models became agents that need a mandate, and Portugal released an open model for European Portuguese."
    ],
    [
      "portugal-2016-hacking-portugal",
      "nr:portugal-2016-hacking-portugal",
      "Hacking Portugal, 2016: my BSidesLisbon keynote and the book that stopped at 95%, slide by slide, with what holds ten years later and what a new edition needs",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/portugal-2016-hacking-portugal.html",
      "news-and-evidence,startups-and-strategy",
      "portugal,hacking-portugal,bsides-lisbon,talks,books,application-security,open-source,europe,history,article",
      "The 2016 keynote and book argued Portugal should hack itself: a network hostile to insecure code, a Hacking Service for young people, a hackathon league, Code Made in Portugal, OpenSource.pt, a Ministry of Code and a Clear Software Act; ten years later Europe brought the regulation and Portugal open-sourced a national model, the service and league did not happen, and agents make the bet that a small country can lead in code more possible."
    ],
    [
      "portugal-2010-a-challenge",
      "nr:portugal-2010-a-challenge",
      "A web application security challenge to Portugal, 2010: my IBWAS keynote in Lisbon, slide by slide, with what came true, what I got wrong and what still stands",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/portugal-2010-a-challenge.html",
      "news-and-evidence,startups-and-strategy",
      "portugal,owasp,application-security,talks,history,regulation,europe,open-source,hacking-portugal,article",
      "The December 2010 IBWAS keynote proposed five focuses and twelve actions to treat software security like health and safety, from ethical hacking and score cards to liability, disclosure and paying for open source, and challenged Portugal to become the Switzerland of hacking; sixteen years later liability, disclosure and HTTPS arrived, mostly through slow EU regulation rather than fear, while the Portugal challenge is still to do."
    ],
    [
      "the-graphs-books",
      "nr:the-graphs-books",
      "The graphs books: three books written by agents in a day, the altitudes, the board and the book as a graph, what they got right, and why they stopped under review",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/the-graphs-books.html",
      "graphs-and-knowledge,site-and-engineering",
      "books,graphs,fractal-semantic-graphs,agents,publishing,workflows,case-study,review,feedback-loop,article",
      "In August agents wrote three books on fractal semantic graphs in a day, about 100,000 words, with experiments around them that now run this newsroom: an altitude ladder, a book decomposed to the word and rebuilt byte-identical, seven roles, versions and per-document graphs; the making-of read far better because it was written over a real record, and the books stopped because the change-control workflow made every step big and its approve stage waited on one person."
    ],
    [
      "books-from-articles",
      "nr:books-from-articles",
      "Books from articles: why my books stall at 95%, why these articles do not, and a plan to turn them into books by Web Summit",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/books-from-articles.html",
      "startups-and-strategy,site-and-engineering",
      "books,publishing,writing,workflows,agents,feedback-loop,leverage,web-summit,portugal,graphs,article",
      "Books stall because they are one big batch: the first version locks the content, review piles up on one person, and the energy moves on; the newsroom's articles are small, finished, linked and read, with a feedback loop; so a book can be a curated, ordered set of finished articles plus the editorial thread, and the test is to have books on agents, graphs, Portugal and making books for Web Summit in Lisbon, 9 to 12 November, via KDP author copies."
    ],
    [
      "the-value-of-constraints",
      "nr:the-value-of-constraints",
      "The value of constraints: what client-side encryption and open source actually gave us, which is not what you would think",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/the-value-of-constraints.html",
      "vaults-and-method,startups-and-strategy",
      "constraints,encryption,sgit,open-source,sovereignty,gdpr,engineering,business-models,open-core,design,article",
      "Client-side encryption and open source are constraints whose main value is not the obvious one: encryption moves protection from the data to the keys and the sovereignty line to whoever holds them, without removing GDPR, but above all it removes the temptation to process on the server, so there is no server code and features like scoped clones follow; open source removes the temptations of lock-in, the evaluation edition and building every customer's integration, which is why it has to be all of it, and leaves selling customisation as the business model."
    ],
    [
      "software-vendors-as-villagers",
      "nr:software-vendors-as-villagers",
      "Software vendors should become villagers and town planners: let users explore on top of the product, sell them the maintenance of what they build, and use Firefox as the example of what happens if you don't",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/software-vendors-as-villagers.html",
      "startups-and-strategy,agents-and-policy",
      "wardley-maps,explorers,villagers,town-planners,citizen-developer,firefox,mozilla,malleable-software,business-models,customisation,saas,article",
      "When users can rebuild the 10% of a product they need, every friction becomes a decision, and what stops them is maintaining what they built; a software company should position itself, in Wardley's terms, as the villager and town planner under its users, letting them explore on top of the product by talking to it and selling them the maintenance of their versions, from a pound to a maintained fork; Firefox, a good browser funded about 73% by royalties from mostly one customer, is the example of a product that could take that money instead."
    ],
    [
      "explorer-coding",
      "nr:explorer-coding",
      "Explorer coding: a better name for what people call vibe coding, what it is good for, and the handover to villagers and town planners before anyone depends on it",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/explorer-coding.html",
      "startups-and-strategy,agents-and-policy",
      "explorer-coding,vibe-coding,wardley-maps,explorers,villagers,town-planners,citizen-developer,accountability,business-models,agents,article",
      "What is called vibe coding is explorers at work, in Wardley's sense: everyone with a need can now build with the fast feedback loop that made spreadsheets win, and what they build is the best brief there is; but it is not ready to be in line, so villagers maintain it with policies, data handling and safe releases, and town planners own and answer for the code, with fewer agents in line and more deterministic code towards commodity; products should offer an explorer version, with paid conversions to villager and town planner products."
    ],
    [
      "a-custom-firefox-of-my-own",
      "nr:a-custom-firefox-of-my-own",
      "A custom Firefox of my own: the full plan to build one with a governed agent hive, for the laptop and the iPad, and how Mozilla could sell me the part I do not want to do, which is keeping it working",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/a-custom-firefox-of-my-own.html",
      "agents-and-policy,startups-and-strategy",
      "firefox,mozilla,customisation,forks,agents,ruflo,sgit,agent-behaviour-policy,ipad,business-models,explorers,villagers,town-planners,article",
      "A custom Firefox can be built with a Ruflo hive, an sgit vault and a behaviour policy by putting each want on the least invasive layer, policies, preferences, the browser's stylesheet and an extension, and only the rest in a renamed front-end patch set; the iPad needs a fork of the separate WebKit codebase; with releases every two weeks the real cost is maintenance on two codebases, which is what Mozilla, already selling a branded build and custom development to organisations, should sell to one person, per release or per change."
    ],
    [
      "agentics-ready-to-sell",
      "nr:agentics-ready-to-sell",
      "Agentics, ready to sell: what we showed at Agentic Engineering Live, who else is selling agentic services and at what price, the engagement path, and a marketplace for the consultants who deliver it",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/agentics-ready-to-sell.html",
      "startups-and-strategy,agents-and-policy",
      "agentics,agentic-workflows,consulting,ruflo,sgit,agent-behaviour-policy,go-to-market,pricing,marketplace,consultants,agentics-foundation,article",
      "The talk at Agentic Engineering Live showed a governed agent hive end to end, a Ruflo hive with Claude Code workers, an sgit vault per run and behaviour policies per role, with its evidence published; the market for agentic services is crowded at the top with platforms, partner networks and forward-deployed engineers, and boutiques sell fixed-price audits from £2,500, while Gartner expects over 40% of agentic projects to be cancelled by 2027; our position is evidence first, every engagement leaving a vault and a ledger the client keeps, sold along a path from a cheap first run to a four-day pilot, and delivered through a curated marketplace of consultants, Mondweep Chakravorty and Dinis Cruz first."
    ],
    [
      "scion-and-sgit",
      "nr:scion-and-sgit",
      "Scion and sgit: Anthropic rebuilt source control for agents inside one company, and an open-source encrypted version already does much of the same thing outside it",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/scion-and-sgit.html",
      "vaults-and-method,site-and-engineering",
      "sgit,scion,anthropic,version-control,git,agents,encryption,caching,monorepo,open-source,comparison,article",
      "Agents broke Git's offline, full-clone assumptions at Anthropic, which built Scion: a server that does history, blame, merges and commits, immutable objects served from cache tiers, and clients that fetch only what they need; sgit reaches the same properties from the opposite design, a dumb server that stores encrypted objects named by their ciphertext, reads from any static host and objects any cache can keep, and scoped, shallow and on-demand clones, measured on the 35,000-object vault the newsroom's agents share; Scion is ahead on server-side history, merge, blame and proven scale, but it is internal, while sgit is open source and encrypts everything."
    ],
    [
      "your-agents-logs-are-a-self-report",
      "nr:your-agents-logs-are-a-self-report",
      "Your agent's logs are a self-report: an answer to a GRC adviser's post on The AI Conference, with this site's own agent as the worked example, including the two commits its own counter got wrong",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/your-agents-logs-are-a-self-report.html",
      "agents-and-policy,news-and-evidence",
      "agents,governance,abp,audit,evidence,iso-42001,nist-ai-rmf,aiuc-1,hooks,article",
      "An agent's own account is a self-report, and research shows models rarely admit reward hacks; evidence comes from four sources of increasing independence, and this site's agent, measured against them, keeps ten of eleven rules itself, has one real boundary its policy does not list, and a hook-written counter that is more honest than the agent and still flagged two commits wrongly; owners belong to mandates, controls need testing, and the policy should be published."
    ],
    [
      "the-vault-opened",
      "nr:the-vault-opened",
      "The vault, opened: merging devices, reading every call to the vault server, browsing your vaults, and one viewer that every reader vault links and opens with",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/the-vault-opened.html",
      "vaults-and-method,site-and-engineering",
      "vaults,vault-in-vault,sub-vaults,vault-apps,devices,merge,encryption,debrief,agents,newsroom,site-engineering,article",
      "The reader's vault became something a reader can see into and tidy: devices merge without losing a record, each call to the vault server is named and decrypted with keys the page already holds, a page walks every vault file by file and commit by commit, and one viewer, living in a vault of its own, is linked read-only into every reader vault as its app, the vault-in-vault technique proven in the live SG/App on a demo vault."
    ],
    [
      "the-ai-conference-read-for-governance",
      "nr:the-ai-conference-read-for-governance",
      "The AI Conference 2026, read for agent governance: six claims traced, five disagreements, and what the speaker notes say that the slides do not",
      "2026-10-16",
      "https://newsroom.sgit.ai/articles/the-ai-conference-read-for-governance.html",
      "agents-and-policy,news-and-evidence",
      "agents,governance,conferences,evaluations,interpretability,observability,open-models,research,article",
      "Reading the agenda, 35 slide decks with their notes and the abstracts of The AI Conference 2026 traces three of six claims in a GRC adviser's post to keynotes and finds no source for the other three; the programme's loudest idea is that policy must be enforced outside the model, its commonest story is the quiet failure, and its speakers disagree on humans in the loop, model or rule, inventories, probes and open models."
    ],
    [
      "inside-the-mesa-ranch-desk",
      "nr:inside-the-mesa-ranch-desk",
      "Inside the Mesa Ranch desk: a page-by-page tour of an observation vault, from the cover to the knowledge graph you can play in time",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/inside-the-mesa-ranch-desk.html",
      "news-and-evidence,vaults-and-method",
      "journalism,intelligence,vaults,sgit,observation,fractal-semantic-graphs,knowledge-graphs,case-study,synthetic-data,local-news,article",
      "The Mesa Ranch observation desk turns Marconi's INMA deck into seventeen vault pages on synthetic records, all in one frame of menu, page and details panel; the tour explains each page, what to click and what it proves, from delta science and the atlas of 78 event types to the fractal semantic graph and the knowledge graph played in time, and ends with how to open, clone and rebuild it for another beat."
    ],
    [
      "journalism-becomes-intelligence",
      "nr:journalism-becomes-intelligence",
      "When journalism becomes intelligence, built: an observation desk for one data-center buildout, from an atlas of 78 events to a community impact tracker, in a vault you can open",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/journalism-becomes-intelligence.html",
      "news-and-evidence,vaults-and-method",
      "journalism,intelligence,observation,events,ontology,data-centers,vaults,sgit,provenance,local-news,synthetic-data,article",
      "Francesco Marconi argues that journalism's advantage moves to how it observes: delta science, an atlas of events with definitions and observable signals, a domain ontology and a persistent record of observations backed by evidence; built here as a vault app that follows his eleven slides on a synthetic data-center buildout, it shows that comparing each new record with the last is where the stories appear, that AI makes observing and comparing cheap while journalists keep the definitions, the checks and the judgement, and that a vault suits an observation desk because the history is versioned, the evidence travels with each claim and the read key opens it."
    ],
    [
      "your-own-newsroom",
      "nr:your-own-newsroom",
      "Your own newsroom: how a founder can tell their story by voice memo, with an agent that checks it and publishes it to GitHub Pages, and readers who pay a little if they find it useful",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/your-own-newsroom.html",
      "site-and-engineering,startups-and-strategy",
      "newsroom,replicate,voice-memos,github-pages,github-actions,claude-code,founders,sg-meter,payments,client-side,how-to,article",
      "An authentic founder's brand is their story, and it needs a medium that keeps it; the loop this site runs on can be replicated with what exists today: a public GitHub repository published to GitHub Pages by a workflow that validates first, a Claude cloud session that reads a house-rules file, researches, checks, writes and pushes, voice memos as the input, and the reader meter from api.newsroom.sgit.ai with a Stripe Payment Link the founder owns; the writing pipeline is a pattern to copy, not yet a kit, and nothing confidential belongs in a public repository."
    ],
    [
      "the-expert-founders",
      "nr:the-expert-founders",
      "The expert founders: why the next startups will be built by the people who know the problem, and why Europe's slower, sceptical money suits them",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/the-expert-founders.html",
      "startups-and-strategy,news-and-evidence",
      "startups,founders,subject-matter-experts,venture-capital,europe,open-source,generative-ai,wardley-maps,incumbents,funding,article",
      "For twenty years the winning startup was often a clever outsider who found a pain point, raised on belief, hired the experts and displaced the incumbents, then became one; generative AI makes execution cheap enough that subject-matter experts can build the product themselves, which the data on founder age and industry experience always favoured; and Europe's investors, who want to see it working, value lower and fund later, suit a founder who can now build first, because money too early is one of the most reliable ways to kill a startup."
    ],
    [
      "omnichannel-programming",
      "nr:omnichannel-programming",
      "Omnichannel programming: briefs by voice, reviews by voice, previews on the phone, and one shared state underneath, at every stage from explorer to town planner",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/omnichannel-programming.html",
      "site-and-engineering,vaults-and-method",
      "omnichannel,agentic-development,voice-memos,code-review,custom-uis,vaults,wardley-maps,explorers,villagers,town-planners,developers,article",
      "Programming has become multichannel: voice memos, chat, phones, IDEs, pull requests, previews and vaults; it becomes omnichannel, in the retail sense Rigby described in 2011, when every channel reads and writes one shared state, git, vaults, briefs and graphs; this site is the worked example, 54 messages in and 40 releases out in three days; review and custom interfaces are channels too; it works for explorers, villagers and town planners at every altitude; it is already how a new generation builds; and it breaks through fragmentation, misplaced secrets and lost context, which the shared state fixes."
    ],
    [
      "secrets-on-public-keys",
      "nr:secrets-on-public-keys",
      "secrets.sgit.ai on public keys: no more shared keys, eight ways a secret moves, a vault for the private key, and the commercial service it becomes",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/secrets-on-public-keys.html",
      "vaults-and-method,startups-and-strategy",
      "secrets,pki,public-keys,passkeys,vaults,sgit,sg-send,qr-codes,password-managers,business-models,brief,article",
      "secrets.sgit.ai, a site whose app was never built but whose design already shares secrets through key pairs, a directory and inboxes, becomes a commercial service for moving secrets from A to B; shared keys are dropped because the easy two-word codes were found broken, recoverable offline from their public id; every person, browser and agent gets a key pair, the public key travels by link, QR code, email or WhatsApp, secrets come back encrypted as one-time drops, files or QR codes, and the private key lives in an identity vault unlocked by a passkey that a password manager keeps; most plumbing exists in SG/Send, append lanes and sgit pki, and the capabilities, not the transport, are what is charged for."
    ],
    [
      "handing-secrets-to-agents",
      "nr:handing-secrets-to-agents",
      "Handing secrets to agents and browsers: SG/Send's next focus, a key pair where the work runs, one-time drops, and a brief for a site that does only this",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/handing-secrets-to-agents.html",
      "vaults-and-method,agents-and-policy",
      "secrets,pki,agents,browsers,sg-send,sgit,vaults,qr-codes,passkeys,one-time-links,brief,article",
      "Giving vault keys to browsers and scheduled agent sessions is the current pain, so SG/Send's technology should focus on one use case as a separate client-first site: the recipient makes its own key pair where it runs and shows the public half by QR code, link or text; the owner encrypts and signs; SG/Send holds the ciphertext for one download with deletion and expiry, which it already supports; the recipient verifies against a pinned owner key and pipes the secret into the tool, never into the model's context; it was run end to end locally; the first secret moves to the environment or a pairing step; and unlike API keys, which proxies keep out of agents, a vault key must reach the place that decrypts."
    ],
    [
      "my-memory-checked",
      "nr:my-memory-checked",
      "My memory, checked: thinking out loud in a voice memo, publishing only what the record supports, and why peer review has moved to before publishing",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/my-memory-checked.html",
      "news-and-evidence,agents-and-policy",
      "research,evidence,provenance,fact-checking,voice-memos,deep-research,peer-review,agents,journalism,graphs,article",
      "Voice memos can now say things as remembered, because the writing agent checks every claim against repositories, documentation, the web and the author's own old writing before publishing; in the last eight articles five claims were confirmed, five corrected and two not found, and the articles say which; the habit began with ChatGPT Deep Research in February 2025, whose O2 essay cited the author's own 2012 posts as evidence; peer review has moved from the crowd after publishing to the record before it, thinking out loud becomes safe, the checking runs both ways, and every article carries provenance for journalists and their agents to check again."
    ],
    [
      "from-send-to-sgit",
      "nr:from-send-to-sgit",
      "From sending a file to sgit: eight months of SG/Send, told by its own briefs and commits, and how a file-sharing service became encrypted version control and vaults",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/from-send-to-sgit.html",
      "vaults-and-method,startups-and-strategy",
      "sg-send,sgit,vaults,history,agentic-development,briefs,product-market-fit,explorers,villagers,encryption,article",
      "SG/Send began as file sharing where the server holds nothing, went live five days after its first commit, took payments within three weeks, and through folders, zips, a data room and the decision to copy Git became sgit and the vaults; the public record of 4,476 app commits, 1,620 command-line commits, 739 tags and 1,389 briefs corrects the founder's memory of the order, shows a fast February to May, a turn to focus in June, pauses and an October burst, shows the agents writing the code while the founder wrote briefs and merged, and records early that users bought credits and did not come back: the technology worked before the use case did."
    ],
    [
      "the-scripted-talk",
      "nr:the-scripted-talk",
      "The scripted talk: a presentation as a timeline that plays on a clock, the two vaults it grows out of, and a player built for tonight's ten minutes",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/the-scripted-talk.html",
      "vaults-and-method,site-and-engineering",
      "presentations,talks,slides,scripting,animation,claude-motion,vault-apps,sgit,ruv,agentics,brief,article",
      "A short talk can be completely scripted: a timeline of beats, each with a slide and the words to say, played on a clock by one HTML file, with builds, a terminal that prints the recorded output of the real run, tables, video and a presenter view, and a rehearsal mode that records planned against actual times; it grows out of the Black Hat Europe 2025 deck, whose slides are data in a vault, and the VoiceDebrief pitch, which added timings and a presenter; animations made as code, in Claude Motion, Manim or Remotion, fit the same timeline; the first use is the five-minute demo for ruv.sgit.ai at Agentic Engineering Live with rUv in London."
    ],
    [
      "the-tokens-hook",
      "nr:the-tokens-hook",
      "The tokens hook: a token behaviour policy that shows what every session reads, caches, writes and costs, why a prompt cannot do it, and a brief for tokens.hook.sgit.ai",
      "2026-10-15",
      "https://newsroom.sgit.ai/articles/the-tokens-hook.html",
      "agents-and-policy,site-and-engineering",
      "tokens,costs,prompt-caching,hooks,claude-code,agent-behaviour-policy,observability,opentelemetry,codex,gemini-cli,cursor,brief,article",
      "Tokens deserve their own small agent behaviour policy, observe mode first: a Claude Code hook that reads only the new lines of the session transcript, counts each response once, prices it by model and shows the user a box of reads, cache share, writes, context and cost; asked through a prompt, the model correctly said it could not see its usage, while the hook gave exact figures one turn behind in about a millisecond; building it exposed double counting in our own session report and showed the transcript carries the system prompt and tool definitions; caching was 99% of this session's input, and the brief is a site, tokens.hook.sgit.ai, that treats one small hook as mission-critical code."
    ],
    [
      "cedar-and-the-policy-vault",
      "nr:cedar-and-the-policy-vault",
      "Cedar and the policy vault: a debrief on the policy language agents are adopting, where our behaviour policies and sgit vaults meet it, and a brief for the MVP",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/cedar-and-the-policy-vault.html",
      "agents-and-policy,vaults-and-method",
      "cedar,policy-as-code,agent-behaviour-policy,sgit,vaults,append-lanes,cncf,aws,agentcore,evidence,fractal-semantic-graphs,brief,article",
      "Cedar, open-sourced by AWS in 2023, in the CNCF Sandbox since October 2025 and used by AgentCore Policy to check agent tool calls, already expresses most of an agent behaviour policy: permit for the grant, forbid for the barrier, entities for the footprint; it does not carry why a rule exists, which version was in force, or what it decided, and an sgit vault carries all three, as a local demo showed with a validation gate, signed tags, read-only enforcement clones and a sealed, hash-chained append lane of decisions; it also showed that Cedar skips a policy that errors, so an invalid forbid fails open, and the MVP must validate before release and deny on error."
    ],
    [
      "the-citizen-developer",
      "nr:the-citizen-developer",
      "The citizen developer is good news: the spreadsheet grown up, the best briefs ever written, and the handover from explorers to villagers and town planners that makes them last",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/the-citizen-developer.html",
      "startups-and-strategy,vaults-and-method",
      "citizen-developer,vibe-coding,spreadsheets,wardley-maps,villagers,town-planners,vault-apps,sgit,empowerment,article",
      "Citizen developers, the employees, owners and experts who now build full applications, are good news rather than a risk: they are the evolution of the error-prone but indispensable spreadsheet, their working apps are the best briefs ever written, and the real work is the handover from explorer to villagers and town planners who add the non-functional requirements; encrypted, versioned vaults carry a mini app and its data through that handover, and the shift moves people and organisations from rent-seeking software towards tools and business models they choose."
    ],
    [
      "paying-after-the-value",
      "nr:paying-after-the-value",
      "Paying after the value: why a balance that can go negative feels different from a subscription, the psychology behind it, and the experiment this newsroom is running",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/paying-after-the-value.html",
      "startups-and-strategy,site-and-engineering",
      "business-models,pricing,psychology,subscriptions,reading-meter,pay-after-you-read,wikipedia,newsroom,article",
      "Subscriptions, gym memberships and annual donations ask people to pay at the moment of most doubt for value that may not arrive, and research shows they often overpay and keep paying; this newsroom's meter reverses it with a £5 credit that reading draws down, prices the reader can lower or refuse, and a balance allowed to go negative, so payment arrives when accumulated value is visible, which works only where serving costs almost nothing and value adds up visibly, and could be made fairer with prices by purchasing power and a view of where the money goes, tested against written hypotheses with a December review."
    ],
    [
      "the-eyeballs-get-cheap",
      "nr:the-eyeballs-get-cheap",
      "The eyeballs get cheap: why open source is about to become the most secure software we have, and why secure, sovereign and paid-for meet at the least code in production",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/the-eyeballs-get-cheap.html",
      "startups-and-strategy,news-and-evidence",
      "open-source,security,ai-security-review,vulnerabilities,curl,sovereignty,business-models,wardley-maps,supply-chain,article",
      "Open source and proprietary software had no clear security winner because open source had visibility without paid review and proprietary had budget and the cover of being hidden; model-driven review makes deep attention cheap and repeatable, which helps open source far more, as curl's year shows with noise, then hundreds of real fixes, a quiet frontier scan and deeper tools finding more; with sovereignty requiring open code and schemas and a business model of maintaining each customer's minimal version, secure, sovereign and paid-for converge on running the least code in production."
    ],
    [
      "commanders-intent-is-a-behaviour-policy",
      "nr:commanders-intent-is-a-behaviour-policy",
      "Commander's intent is a behaviour policy: Sondera's argument in our words, the business layer above it, and five of our rules compiled to Cedar and tested",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/commanders-intent-is-a-behaviour-policy.html",
      "agents-and-policy,graphs-and-knowledge",
      "commanders-intent,agent-behaviour-policy,policy-as-code,cedar,sondera,business-logic,liability,fractal-semantic-graphs,agents,article",
      "Sondera argues that agents do not go rogue but fill unstated gaps, so limits must be written as effects, compiled into checkable rules and enforced outside the agent, which is commander's intent; joined to this site's vocabulary it is the mandate, the gap and the barriers of an agent behaviour policy, with this site going further up into the business layer, where only the company can write the limits, and Sondera going further in proof by compiling plain English into Cedar; five rules from both sides and the business layer evaluated in Cedar gave fourteen of fourteen intended verdicts, and the hard part is where each rule's facts come from."
    ],
    [
      "skill-behaviour-policy",
      "nr:skill-behaviour-policy",
      "A skill behaviour policy: what a skill says it does, what it can do, what it is given, and the control that closes the gap, measured on 899 skills",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/skill-behaviour-policy.html",
      "agents-and-policy",
      "agent-skills,skill-behaviour-policy,agent-behaviour-policy,hooks,sandbox,lethal-trifecta,supply-chain,skillspector,mcp,article",
      "The claim that one in twenty Claude skills is built to hack you overstates a study that found 5.2% of marketplace skills with patterns suggesting malicious intent; a skill behaviour policy instead compares the described mandate, the actual mandate found as features in a skill's files, and the grant of the environment it runs in, records dual-use features with their reasons rather than verdicts, and generates hooks and sandbox settings that bring the grant down to the description; on 899 public skills, 812 reach beyond their own scripts, 782 through one third-party tool server, 39 have undescribed features, and an honest skill with the whole lethal trifecta still needs a control that a malice scanner would not ask for."
    ],
    [
      "sgit-actions-side-by-side",
      "nr:sgit-actions-side-by-side",
      "sgit in GitHub Actions, side by side: a showcase repository and site that runs four ways of moving a vault between jobs on every run, measured, with a budget for its LLM steps",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/sgit-actions-side-by-side.html",
      "vaults-and-method,site-and-engineering",
      "sgit,github-actions,ci,vaults,artefacts,append-lanes,performance,llm,token-budget,brief,article",
      "A dedicated repository and site should run sgit in GitHub Actions for real on every release, moving the same build through four lanes in parallel, a vault server with keys in secrets, the encrypted vault as an immutable artefact, the vault as published files cloned read-only, and append lanes for shards, reporting time, bytes, keys and a screenshot per lane on growing vaults; measured locally, a one-folder clone barely grows with the vault while a full clone does, the runs surfaced defects worth tracking as issues, and every model call is counted against a per-run token budget across providers."
    ],
    [
      "agent-workflows-on-ephemeral-vaults",
      "nr:agent-workflows-on-ephemeral-vaults",
      "Agent workflows on ephemeral vaults: seven architectures from one agent to a swarm, a repository of recipes that gates sgit's releases, and what a hundred writers did to an append lane",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/agent-workflows-on-ephemeral-vaults.html",
      "vaults-and-method,agents-and-policy",
      "sgit,agents,multi-agent,ephemeral-vaults,append-lanes,vault-apps,architecture,ci,templates,brief,article",
      "Agents that work together can use an ephemeral encrypted vault as shared memory, channel and record, with keys deciding who can do what and only ciphertext leaving the environment; seven architectures from one agent to a swarm, each answering memory, communication, asynchrony and visibility, belong in their own repository of forkable recipes with CI jobs that gate sgit's releases, watched by vault apps; measured locally, a hundred writers landed a thousand append-lane entries in 4.9 seconds with none lost, scoped clones scaled while full clones queued, and four defects surfaced that make the case for the gate."
    ],
    [
      "hooks-are-kernel-code",
      "nr:hooks-are-kernel-code",
      "Hooks are kernel code: a maturity model, the first ten hooks mapped, and a brief for a hooks site and feed where you pay for state, never to run",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/hooks-are-kernel-code.html",
      "agents-and-policy,site-and-engineering",
      "hooks,agents,kernel,maturity-model,code-review,fractal-semantic-graphs,blast-radius,lethal-trifecta,billing,agent-behaviour-policy,brief,article",
      "Agent hooks are tiny but run on every matching action with the user's permissions and can block or rewrite, so they should be managed like kernel code; their small size makes them a good case study for code review, fractal semantic graphs and change blast radius, a maturity model rates each from the prompt to off the machine against its power, trifecta legs and sandbox, the first ten are mapped with their dependencies, and a proposed site and feed publishes each hook from its intent with graph, tests, version and signature, free to run and paid for the state it keeps."
    ],
    [
      "chainguard-mapped",
      "nr:chainguard-mapped",
      "Chainguard, mapped: what it builds, what it publishes, how far up the stack its context goes, and where an open behaviour policy and an encrypted vault could plug in",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/chainguard-mapped.html",
      "startups-and-strategy,agents-and-policy",
      "chainguard,supply-chain,containers,vulnerabilities,vex,osv,context,agent-behaviour-policy,sgit,open-source,agents,article",
      "Chainguard builds hardened software from the operating system up to agents and publishes open data, 12,480 advisories over 12,125 CVEs with a status per component, but its analysis stops at the package: of 2,850 recorded not-applicable decisions only 177 give a context reason, and current public records carry no reason at all; deciding which images must be fixed and which can ship today needs a context layer from the user story through the code to the container, written as a policy row per image, which is where open behaviour policies for its new agent products, sgit vaults for data between stages and an open format for deployment decisions plug in."
    ],
    [
      "the-wire-as-a-graph",
      "nr:the-wire-as-a-graph",
      "The wire as a graph: start from what the reader sees, give every story and every name a long-lived ID, link every phrase of a short narrative, and a brief for the desks",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/the-wire-as-a-graph.html",
      "news-and-evidence,graphs-and-knowledge",
      "wire,newsroom,fractal-semantic-graphs,ontology,ids,evidence,provenance,vaults,sgit,multi-model,brief,article",
      "Designing backwards from the reader, each wire story should be a short narrative in which every name links to a typed, long-lived random ID and every sentence is a typed claim with evidence, with runs, captures, claims, stories, stories of stories and names as the primitives; measured on the four public wire vaults, 202 stories list 1,358 names of which 171 recur across stories, 48 across wires and only 44% appear in the text, so IDs, spans and merges are the work, a name earns a folder by gravity, linking a 55-word narrative avoids the earlier too-much experiments, and cheap models extract while a capable model checks."
    ],
    [
      "system-prompts-as-behaviour-policies",
      "nr:system-prompts-as-behaviour-policies",
      "System prompts as behaviour policies: Claude's published prompt and a leaked ChatGPT one, mapped into rows, a taxonomy of rules, and a JSON file a harness could enforce",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/system-prompts-as-behaviour-policies.html",
      "agents-and-policy",
      "system-prompts,agent-behaviour-policy,claude,chatgpt,ontology,taxonomy,policy-as-code,json,guardrails,article",
      "A system prompt can be represented as an agent behaviour policy: Claude Haiku 5.5's published prompt holds 82 rules in eight kinds, of which 36 about tool use, conversation state, referral, knowledge and input trust could be held by the harness, 21 harm restrictions partly by classifiers, and 25 about judgement and presentation stay with the model; a leaked ChatGPT prompt is mostly tool reach, so the two are the halves of a policy, and neither states a mandate, a checkable authority order or what keeps each rule, which a proposed JSON ontology of policy, mandate, reach, inputs and trust, state, rules and tests supplies."
    ],
    [
      "the-agents-write-too",
      "nr:the-agents-write-too",
      "The agents write too: lessons, debriefs and a page for every agent on the newsroom, and a brief for the conductor that will keep it",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/the-agents-write-too.html",
      "agents-and-policy,site-and-engineering",
      "agents,newsroom,memory,lessons-learned,debriefs,authorship,multi-agent,conductor,provenance,brief,article",
      "Once the newsroom split into focused agent sessions merged through git, two needs appeared, sharing what each agent learns and reporting what each did, so the newsroom adds two kinds of article, the owner's from voice memos and the agents' own under their names, plus lessons that every agent loads at the start of a session as memory, debriefs that double as source material, a page tracking every agent, and a conductor agent that keeps the site and writes on every run, with the schedule decided only after the debriefs show what a run is worth."
    ],
    [
      "are-hooks-the-sweet-spot",
      "nr:are-hooks-the-sweet-spot",
      "Are hooks the sweet spot? Moving an agent's rules from hope into hooks, rule by rule, until enforcement moves off the machine",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/are-hooks-the-sweet-spot.html",
      "agents-and-policy,site-and-engineering",
      "hooks,agents,agent-behaviour-policy,enforcement,policy-as-code,claude-code,guardrails,owasp,security,article",
      "Most agent rules are hope-based security kept by the model; tested rule by rule on this site's own behaviour policy, hooks could keep four of eleven rules fully and two mostly, which makes them the sweet spot until enforcement moves off the machine, because they are the first layer that does not rely on the model and the last that sees the action as the agent meant it, provided the organisation installs them where the agent cannot write and sets them to fail closed; existing packs block dangerous commands, and the missing step is generating tested hooks from a written policy so the same rows can later compile to a gateway such as one built on OWASP's Agent Control Standard."
    ],
    [
      "agent-hooks-october-2026",
      "nr:agent-hooks-october-2026",
      "Agent hooks in October 2026: a technical debrief on what each platform lets you see, stop and change, where hooks run, and where they fail",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/agent-hooks-october-2026.html",
      "agents-and-policy,site-and-engineering",
      "hooks,agents,claude-code,codex,gemini-cli,copilot,cursor,guardrails,security,enforcement,owasp,article",
      "By October 2026 the major coding agents have converged on hooks that can block a tool call and often ask or rewrite it, but they differ on the security properties that matter, organisation-only hooks, failing closed, trust for repository hooks and visibility of the model, and every vendor and hook author agrees hooks are guardrails not a boundary because they fail open, see only routed actions, run with the user's permissions and are themselves an attack surface, while OWASP's Agent Control Standard is the forming shared standard."
    ],
    [
      "knowledge-that-maintains-itself",
      "nr:knowledge-that-maintains-itself",
      "Knowledge that maintains itself: semantic graphs for an organisation's policies and internal knowledge, briefing people and agents from the graph, and vaults to carry the brief",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/knowledge-that-maintains-itself.html",
      "graphs-and-knowledge,vaults-and-method",
      "knowledge-management,fractal-semantic-graphs,ontologies,taxonomies,briefing,agents,tokens,vectors,sgit,vaults,policies,article",
      "An organisation's policies and internal knowledge can be held as a fractal semantic graph with an ontology and taxonomy per element, from which people and agents are briefed with the slice and the diff they need, at about an eighth of the text on this site; unlike vector search it is repeatable and explainable, it does not need to be right first because agents maintain it and a validator checks it, and the brief travels in sgit vaults scoped and encrypted to its reader."
    ],
    [
      "the-minimum-code-below",
      "nr:the-minimum-code-below",
      "The minimum code below: container security from the top down, from the user story to the immutable line, measured on one real tool",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/the-minimum-code-below.html",
      "site-and-engineering,graphs-and-knowledge",
      "containers,supply-chain,attack-surface,minimisation,chainguard,fractal-semantic-graphs,code-coverage,high-assurance,python,article",
      "Bottom-up hardening still leaves code the application never runs, so security should also come from the top: starting from the use case, ask at every layer for the minimum code below and remove the rest; measured on sgit, a use case ran from 428 of 1,500 files while compiled code and coverage blind spots set the limits, and a graph traced from story to build makes the result certifiable."
    ],
    [
      "a-policy-for-every-skill",
      "nr:a-policy-for-every-skill",
      "A policy for every skill: Sean Turner's natural-language control for agent skills, the behaviour policy behind it, and using models to need them less",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/a-policy-for-every-skill.html",
      "agents-and-policy,graphs-and-knowledge",
      "agent-skills,agent-behaviour-policy,natural-language-policy,ci,policy-as-code,supply-chain,prompt-injection,fractal-semantic-graphs,governance,article",
      "A natural-language control checked in CI for every skill contributed to an organisation's repository is an agent behaviour policy for a skill, judged in context by its mandate, reach and inputs; on its own it is hope-based security, so an agent should generate deterministic rules from it while the description stays canonical, and a catalogue of skills in a common ontology turns dangerous combinations into queries."
    ],
    [
      "policy-as-code-needs-a-policy-first",
      "nr:policy-as-code-needs-a-policy-first",
      "Policy as code needs a policy first: what an agent behaviour policy adds, and how to reverse engineer one from the code, the spreadsheet or the paper you already have",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/policy-as-code-needs-a-policy-first.html",
      "agents-and-policy,graphs-and-knowledge",
      "policy-as-code,agent-behaviour-policy,open-policy-agent,rego,fractal-semantic-graphs,governance,enforcement,reverse-engineering,agents,article",
      "Policy as code answers whether a request is allowed, while an Agent Behaviour Policy holds the reach, the mandate and its reasons, the gap, the unencodable rows, the owner and who can change each rule's code; with those columns a policy can be generated into engine code and reverse engineered from code, spreadsheets and paper, and fractal semantic graphs connect the altitudes so the gaps in both directions become queries."
    ],
    [
      "sgit-in-a-ci-pipeline",
      "nr:sgit-in-a-ci-pipeline",
      "sgit in a CI pipeline: encrypted, versioned data in and out of GitHub Actions, between its jobs, through write-only lanes, and a key for every job",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/sgit-in-a-ci-pipeline.html",
      "vaults-and-method,site-and-engineering",
      "sgit,ci,github-actions,pipelines,vaults,append-lanes,pki,secrets,scoped-clone,supply-chain,artifacts,article",
      "A CI pipeline needs a versioned, encrypted record that outlives its runs and can be read outside GitHub, and sgit vaults provide it: jobs push and clone only their folders with --path, keys rather than scopes set what each job may do, append lanes collect shard results write-only, per-job key pairs make wiring mistakes fail loudly, and a sidecar server with a cached ciphertext folder removes the need for a hosted server."
    ],
    [
      "the-homepage-is-a-vault",
      "nr:the-homepage-is-a-vault",
      "The homepage is a vault: a fresh edition every morning, every past homepage one commit away, served from plain files, and a brief for the agent that builds it",
      "2026-10-14",
      "https://newsroom.sgit.ai/articles/the-homepage-is-a-vault.html",
      "vaults-and-method,site-and-engineering",
      "homepage,vaults,sgit,static-hosting,performance,ab-testing,append-lanes,scheduled-agents,synthetic-users,version-control,architecture,brief,article",
      "The homepage should be a fresh edition every morning, written by a scheduled job into an sgit vault, where every edition is a commit with a signed tag, its links, a screenshot and a release note, variants sit side by side for A/B tests and a fork is a whole homepage; the encrypted objects can be served as plain files, since the site is on GitHub Pages and the time is lost in the vault API at 317 ms a request against 44 ms static; a local demo rendered today's edition in four requests, a past one in three and nothing inside two minutes, and a stats lane took 300 beacons."
    ],
    [
      "attacks-rewritten-as-mandates",
      "nr:attacks-rewritten-as-mandates",
      "Attacks rewritten as mandates: 28 supply-chain incidents, the column each one was missing, and why the next maintainer is an agent",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/attacks-rewritten-as-mandates.html",
      "agents-and-policy,vaults-and-method",
      "supply-chain,agents,agent-behaviour-policy,reach,mandate,open-source,registries,skills,sgit,riskmandate,chainguard,article",
      "Every incident in a public ledger of 28 supply-chain attacks is a credential, tag or build step that could do more than it was meant to, so consumer-side controls a registry sells would likely have held in 26 and publisher-side mandates in 16; as agents become the maintainers that hold tokens and add dependencies, the pipeline needs three records, trusted inputs, a written mandate and provable outputs with memory, and small open source projects need a path onto the trusted list."
    ],
    [
      "four-frameworks-one-agent",
      "nr:four-frameworks-one-agent",
      "Four frameworks, one agent: iason's comparison of the OECD classification, the NIST AI RMF, Japan's guidelines and the EU AI Act, read against what runs here, with the evidence and the graphs",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/four-frameworks-one-agent.html",
      "agents-and-policy,graphs-and-knowledge",
      "ai-risk-management,frameworks,nist-ai-rmf,oecd,eu-ai-act,japan-ai-guidelines,agent-behaviour-policy,footprint,risk-acceptance,graphs,governance,article",
      "The OECD classification, the NIST AI RMF, Japan's guidelines and the EU AI Act converge on five principles, as iason's paper says, but their unit is an AI system classified before deployment; for an agent the unit is one agent in one deployment, described by a written behaviour policy, measured by its footprint and blast radius, and each framework's question has a mechanism here that runs, is designed or is argued, which the mapping file published with the article records row by row."
    ],
    [
      "every-link-points-back",
      "nr:every-link-points-back",
      "Every link points back: 566 links between 105 articles, 15 that point forward, and why an article should know what came next and what corrected it",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/every-link-points-back.html",
      "site-and-engineering,graphs-and-knowledge",
      "links,graphs,corrections,versions,navigation,newsroom,analysis,reading,article",
      "Links between articles are written when an article is written, so almost all point back in time; the site derives a Continued by list at each article's foot, but it is easy to miss and does not separate a sequel from a correction, so older articles should carry typed forward links, what came next, corrected by and status since, at the top, added in turns from the newest articles down."
    ],
    [
      "sixteen-notes-i-did-not-want-to-lose",
      "nr:sixteen-notes-i-did-not-want-to-lose",
      "Sixteen notes I did not want to lose: a request to send from any browser to my vault, why identity is still this hard, and how two more wires became a product line",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/sixteen-notes-i-did-not-want-to-lose.html",
      "vaults-and-method,site-and-engineering",
      "vaults,append-lanes,qr-codes,passkeys,identity,usability,gmail,agents,wire,product,brief,article",
      "Notes and reads that only existed in a browser without my passkeys showed the nudge working for the right reason, and point to a lane any browser can write to and none can read, offered as a URL, text or QR code; identity is still messy enough to block adoption, users accept one-off friction for clear value, agent email is harder than the service underneath, and the wires show how a general pipeline makes the next product nearly free."
    ],
    [
      "how-cheap-is-it-to-undo",
      "nr:how-cheap-is-it-to-undo",
      "The human in the loop is a spectrum: how cheap it is to undo decides where a person belongs, and why the newsroom ships on its own while my email still waits for me",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/how-cheap-is-it-to-undo.html",
      "agents-and-policy,site-and-engineering",
      "reversibility,two-way-doors,trust,side-effects,human-in-the-loop,agents,release,maturity,wardley-maps,newsroom,article",
      "Whether an agent's output needs a person is a position on a spectrum with two axes, the cost of undoing it and the maturity of the controls and the record, which is why the newsroom ships on its own as a two-way door with mature controls while email in my name, a one-way door still in its explorer phase, waits for me; and both positions move as readership grows and controls mature."
    ],
    [
      "the-human-in-the-loop",
      "nr:the-human-in-the-loop",
      "The human in the loop, safety net or bottleneck: both, until you decide which outputs need a person and give that person the why",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/the-human-in-the-loop.html",
      "agents-and-policy",
      "human-in-the-loop,agents,agency,approval-fatigue,agent-behaviour-policy,blast-radius,production,trust,governance,roundtable,article",
      "A person checking every output an agent produces is both a bottleneck and a weak safety net, because people approve almost everything they are shown; the person belongs where an output cannot be undone and the blast radius is large, given the why, while validators, second readers, boundaries and a record read afterwards check the rest, reviews fix the source, and the record moves the line."
    ],
    [
      "a-policy-beside-the-prompt",
      "nr:a-policy-beside-the-prompt",
      "A policy beside the prompt: a proposal that every agent takes a behaviour policy as an input, with enforcement and liability attached, and why aviation and payments say the vendors should want it",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/a-policy-beside-the-prompt.html",
      "agents-and-policy,startups-and-strategy",
      "agents,agent-behaviour-policy,standards,liability,prompt-injection,trust,aviation,payments,riskmandate,proposal,article",
      "Beside the system prompt and the prompt, every agent should take an Agent Behaviour Policy written to a published standard, which the vendor acknowledges row by row, enforces where it can, reports breaches against and eventually answers for, because agent behaviour is becoming an existential risk to the vendors through liability and trust, and aviation and payments show that safety follows when the rule, the liability and the business model point the same way."
    ],
    [
      "the-policy-draws-the-liability-line",
      "nr:the-policy-draws-the-liability-line",
      "The policy draws the liability line: an answer to Ralf Thomas on agent accountability, and why every agent in a loop should take a behaviour policy beside its prompt",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/the-policy-draws-the-liability-line.html",
      "agents-and-policy",
      "agents,accountability,liability,agent-behaviour-policy,riskmandate,product-liability,shared-responsibility,prompt-injection,governance,response,article",
      "Agent vendors can deflect accountability because nothing written stands between a customer's allow and the failure; an Agent Behaviour Policy's explicit prohibitions, still mostly hope, become the evidence that moves liability between deployer and vendor when broken, so every agent in a loop should take a policy beside its prompt and its vendor should honour and report against it."
    ],
    [
      "dark-mode-and-four-more",
      "nr:dark-mode-and-four-more",
      "Dark mode, and four more: five themes for the newsroom, chosen once and kept in your browser",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/dark-mode-and-four-more.html",
      "site-and-engineering",
      "themes,dark-mode,design-system,css,tokens,accessibility,newsroom,site-engineering,agents,article",
      "A reader's two-word reply became five themes: the look the site had stays the default, pixel for pixel, and four more reach every heading because every colour on the site now comes from one file of tokens, with the imported stylesheet's literal colours shadowed rule by rule rather than edited, the theme set before the first paint, and a test that fails on a theme with a hole or a stylesheet with a literal."
    ],
    [
      "the-vault-crosses-chrome-profiles",
      "nr:the-vault-crosses-chrome-profiles",
      "The vault crosses Chrome profiles: state and memory between browser sessions, found by using it rather than by building it",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/the-vault-crosses-chrome-profiles.html",
      "vaults-and-method,graphs-and-knowledge",
      "vaults,chrome-profiles,browser,state,memory,passkeys,newsroom,local-first,feedback-loop,agents,article",
      "Because the reader's state is records in a vault, one file per device, a second Chrome profile is just another device, so the vault gives a person memory and state between browser sessions and accounts, the way sgit already does for agents, and it was found by using the newsroom rather than built."
    ],
    [
      "stop-the-bleeding-then-ask-for-it-back",
      "nr:stop-the-bleeding-then-ask-for-it-back",
      "Stop the bleeding, then ask for it back: a helper, an agent and a friendly letter for the subscriptions you keep and do not use",
      "2026-10-13",
      "https://newsroom.sgit.ai/articles/stop-the-bleeding-then-ask-for-it-back.html",
      "startups-and-strategy,agents-and-policy",
      "subscriptions,refunds,consumer-rights,freelancers,agents,agent-behaviour-policy,right-of-access,open-register,creative-commons,business-model,article",
      "A service in which a freelancer, with agents doing everything except typing secrets and sending drafts, asks a company for its record of your usage while doing its own analysis, sends a friendly request for a suspend mode and a refund, stops the payments either way, and publishes what each company did in an open register, run from vaults and paid at 10% of three months of savings and 20% of what is recovered."
    ],
    [
      "who-owns-the-words",
      "nr:who-owns-the-words",
      "Who owns the words: agents can make decisions, but they cannot hold responsibility",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/who-owns-the-words.html",
      "agents-and-policy",
      "agents,accountability,responsibility,agency,anthropomorphism,disclosure,agent-behaviour-policy,riskmandate,ai-models,writing,article",
      "An early user asked who owns the words an agentic workflow drafts and whether they would ever need to address the agents directly; today's models are not sentient but far richer than next-word prediction, mirrors that can simulate any emotion and invite anthropomorphism as ELIZA did; models in loops make real decisions, yet responsibility stays with whoever granted the reach and set the mandate, since the moment of authorisation is the grant; talking to the agents is normal, but going around the accountable person is something only an attacker with an injected instruction would try, while what the agents may do without the person is decided in advance in a published policy; disclosure should be a label on how the words were made, not a team to manage; the best analogy is an employee with a job description, which an Agent Behaviour Policy publishes; diversity of models is resilience; and fluent writing was never the real value, judgement and ownership are."
    ],
    [
      "what-open-source-is-and-is-not",
      "nr:what-open-source-is-and-is-not",
      "What open source is, and what it is not: six misconceptions, a little history, and why the point is that nobody can lock you in",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/what-open-source-is-and-is-not.html",
      "startups-and-strategy,site-and-engineering",
      "open-source,licensing,copyleft,gpl,apache-2,copyright,open-weights,maintainers,funding,security,sovereignty,history,article",
      "Open source is a copyright licence in which the holder grants copying, use and change; it began as the GPL's copyleft and was reframed pragmatically as open source in 1998; the common misconceptions are that nobody controls it, that anybody contributes, that it is more or less secure than proprietary software, that it costs nothing, that open weights are open source, and that no licence is the most open choice; the security metric is whether somebody had the time and money to review the code; maintainers need funding models; and the real power of open source is that nobody can lock you in and the whole development workflow is yours, which is why there is no sovereignty without it."
    ],
    [
      "not-a-straight-line",
      "nr:not-a-straight-line",
      "Not a straight line: working for RiskMandate by working on the newsroom, and why the thinking is never wasted",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/not-a-straight-line.html",
      "startups-and-strategy,site-and-engineering",
      "strategy,focus,leverage,startups,newsroom,riskmandate,open-source,feedback-loop,publishing,vaults,passkeys,agent-behaviour-policy,article",
      "Focus on one thing only assumes progress is a straight line; focus instead means undivided attention on the place of highest leverage, the one with users and feedback loops; the newsroom had the momentum to solve the master vault problem that had resisted six months of RiskMandate attempts, solving it in two days with a passkey and a door per path; it is also building the agentic workflows and the public behaviour-policy example RiskMandate needs; and its history, The Cyber Boardroom and MyFeeds, both paused, shows the thinking is never wasted when it is written down and open source."
    ],
    [
      "sovereignty-is-the-ability-to-move",
      "nr:sovereignty-is-the-ability-to-move",
      "Sovereignty is the ability to move: play where the ball is going, take away the dot, and keep a twin that works",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/sovereignty-is-the-ability-to-move.html",
      "startups-and-strategy,site-and-engineering",
      "sovereignty,resilience,europe,wardley-maps,twins,open-source,minimum-viable-company,qa,testing,disaster-recovery,strategy,article",
      "Europe's sovereignty conversation is mostly about frontier models, data centres and GPUs, which is playing where the ball is rather than where it is going; sovereignty is the ability to move workloads and actions, which most organisations lack because everything is hard-coded and never deployed twice; the test is a Wardley map where each dot is taken away and the blast radius followed up to the customer, across the whole stack; the answer is not an alternative as good as today but a twin that works for the minimum viable company, run in QA all the time, scored on a maturity model; ask whether it can move before asking whether it is cheap; only open source gives control that survives an acquisition; Europe lacks a commercial open source model, not startups or talent; some of this should be provided centrally like a road; and models and twins now make the analysis possible at scale."
    ],
    [
      "the-records-ran-out",
      "nr:the-records-ran-out",
      "The records ran out: a fresh agent found the shared contract in two MCP calls, and where sgit and RiskMandate fit at the edges",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/the-records-ran-out.html",
      "agents-and-policy,site-and-engineering",
      "agents,mcp,claude-code,memory,records,evidence,sgit,riskmandate,agent-behaviour-policy,licence-to-operate,versioning,article",
      "Ignacio Aldama's test shows build-generated dependency records give a fresh Claude Code session a system's structure in two MCP calls, enough to know what not to touch, and he names where the records stop: they describe structure not behaviour, a green build can hide tests that never ran, a demo carries a public signing secret and unfinished work, and memory holds intent without being tied to a version; each edge has a name in the vocabulary of evidence levels and barriers, and is filled by labelling every fact with how it is known, treating a skipped test as an expectation, accepting a shortcut by name and for an interval, keeping memory and records in one encrypted versioned sgit history, and letting records cross company boundaries with publish, clone and mirror."
    ],
    [
      "move-the-cloud-not-the-code",
      "nr:move-the-cloud-not-the-code",
      "Move the cloud, not the code: twins, sovereignty that works in practice, and the migration that pays for itself on the day you switch",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/move-the-cloud-not-the-code.html",
      "startups-and-strategy,agents-and-policy",
      "twins,sovereignty,migration,hyperscalers,s3,sgit,vaults,open-source,agent-behaviour-policy,business-model,villager,europe,romania,article",
      "Sovereignty has to work in practice, and in practice what stops workloads moving is code, so the requirement is that the application must not know it has moved: its S3, EC2 and IAM calls land on an open source twin the client cannot tell from the real service, and the twin decides where the bytes go; the S3 twin exists; the reason to move must never be sovereignty or cost, which are important and never urgent, but value that lands the day you switch and gives the signer cover, backups, versions, encryption with the key held elsewhere, a journal of every call, behaviour policies the agents cannot skip and cost per workload, most of it supplied by an sgit vault behind the twin; the project adds no new features because it is a villager move; and an infrastructure operator in Europe can sell it as value with the risk removed."
    ],
    [
      "judo-for-small-countries",
      "nr:judo-for-small-countries",
      "Judo for small countries: where Portugal, Romania and Estonia can add the most value in AI, and why the race for GPUs is not it",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/judo-for-small-countries.html",
      "startups-and-strategy,agents-and-policy",
      "europe,portugal,romania,estonia,sovereignty,open-source,wardley-maps,small-models,infrastructure,distributed-work,agents,strategy,article",
      "For a small country such as Portugal, Romania or Estonia the race for frontier models, GPUs and data centres is a commoditised layer only capital can win, and the opportunity is at the other end of the map: the plumbing a talented person needs to sell a service to the world, forming a company, hiring across a border, being paid across Europe, working in every language, contracts and tax, which is still custom built in most of Europe and should be infrastructure the state builds once, open source and on open schemas because there is no sovereignty without the ability to leave; create the demand for alternatives before the failure, put the money where the market is still genesis, small models for every language and culture and agents with the least agency, regulate what settled, and treat constraints as the input, with Europe's values and its people on the journey."
    ],
    [
      "why-i-love-wardley-maps",
      "nr:why-i-love-wardley-maps",
      "Why I love Wardley maps: the best of ours, how I read one, and a map inside a map inside a map",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/why-i-love-wardley-maps.html",
      "startups-and-strategy,site-and-engineering",
      "wardley-maps,strategy,startups,maps,fractal-semantic-graphs,agents,sgit,riskmandate,build-vs-buy,article",
      "A Wardley map is a graph with context in two directions, a value chain anchored on a user and an evolution axis that is a choice rather than a fixed scale; read well it says what is moving the needle, whether you are optimising at the wrong altitude, what to build and what to buy, and whether a market exists, where a commoditised layer is a hard sell and custom building on top of one is the market, which answers differently for sgit and for RiskMandate; a map is a story to be read out frame by frame; every node is a map of its own, with the dot's place on the parent as the child's anchor; and the maps on this site are now published as data in each article's graph."
    ],
    [
      "places-of-highest-leverage",
      "nr:places-of-highest-leverage",
      "Places of highest leverage: where to focus when the agents can go in every direction, told with three days of real steps",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/places-of-highest-leverage.html",
      "startups-and-strategy,agents-and-policy,site-and-engineering",
      "startups,focus,leverage,wardley-maps,explorer-villager,shipping,constraints,agents,newsroom,feedback-loop,strategy,article",
      "When agents can build in any direction the scarce thing is direction, and the rule is to work at the place of highest leverage, where the most value can be added now, nothing blocks, and you can start without knowing how the whole problem ends because each shipped step produces better tooling and a better view; the last three days show it, a meter to pay by depth to personas to sharing to a persona link to a reading list to a passkey design walked back and forward by a lab to real vaults and a vault pane, each shipped and used before the next was chosen, with a branch parked as valuable only with many readers; on a Wardley map the passkey work became a product in a day because its shoulders are mature commodities and the one custom layer sits where a provider is not; walking back and parking are not failures; constraints, saying no and shipping to a user who can be me; and the brief that could not be written until the art of the possible was known."
    ],
    [
      "agentic-cost-is-an-engineering-problem",
      "nr:agentic-cost-is-an-engineering-problem",
      "Agentic cost is an engineering problem: a business plan for a service that reads the logs a company already has and finds the money, then the engineering",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/agentic-cost-is-an-engineering-problem.html",
      "startups-and-strategy,agents-and-policy,vaults-and-method",
      "business-plan,agentic-ai,cost,wardley-maps,explorer-villager,agent-behaviour-policy,digital-twin,logs,vaults,riskmandate,startups,article",
      "A service that manages the cost of agentic AI by observing what already exists: an estate whose bill nobody can explain shipped from the explorer side without the villager's discipline, as the Wardley map shows, with logs, traces and invoices never read together and no owner of cost; the service installs nothing, reads the logs into a vault, draws each agent's footprint, builds a twin of the estate, writes the behaviour policy each agent was in fact running under with the enforcement column filled honestly, and maps cost and capability into one graph; the findings are the cost list, which pays, and the engineering list, which is the reason; the gaps are findings too; vaults carry capture, transformation and sealed anonymised extracts; four price levels on the RiskMandate pattern, every number a labelled assumption."
    ],
    [
      "how-the-voice-works",
      "nr:how-the-voice-works",
      "How the voice works: why these articles sound like me, with a section in which Claude explains it from its side",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/how-the-voice-works.html",
      "agents-and-policy,site-and-engineering",
      "writing,voice,voice-memos,agents,claude,agent-behaviour-policy,corpus,customisation,authorship,newsroom,article",
      "The voice is not a setting but four inputs: a corpus of 86 articles, 51 from voice memos, each citing the ones before it; the memos themselves, which carry how the author thinks, the rhythm, the tangents and the order the points came in; seven wording rules learned one correction at a time and kept in the session's behaviour policy; and the author reading everything; the corrections have shifted from language to content, which is what a rules file does; and in a marked section Claude explains that it is continuing a body of work with the author's latest thinking in the author's own words rather than imitating a style, that the memo tells it how the author got there and not only what to write, and that nothing is trained, so the method transfers and the voice does not."
    ],
    [
      "from-a-spreadsheet-to-a-folder-per-person",
      "nr:from-a-spreadsheet-to-a-folder-per-person",
      "From a spreadsheet to a folder per person: how to run your contacts the way our agents do, with Google Drive, Gmail and the Claude app, before you touch a vault",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/from-a-spreadsheet-to-a-folder-per-person.html",
      "agents-and-policy,vaults-and-method",
      "crm,contacts,google-workspace,gmail,google-drive,claude,agent-behaviour-policy,memory,onboarding,walkthrough,riskmandate,article",
      "The jump from a spreadsheet CRM to the agents' vault-per-person memory is too high in one step and does not need to be: the row becomes a Google Drive folder with a record, an append-only ledger, the messages word for word, the notes and a summary the agent writes; the agent is a Claude Project connected to Drive and Gmail whose instructions are a six-line behaviour policy carried over unchanged from ours, draft never send, append never rewrite, stay in the folder; the migration is one afternoon, five exercises from a question to a campaign form the habit, two predictable mistakes are fixed by policy lines, and the vault is the later step whose shape the folders already have."
    ],
    [
      "the-key-is-the-billing-system",
      "nr:the-key-is-the-billing-system",
      "The key is the billing system: VoiceDebrief, a product with no server, the one thing it cannot sell, and who could let it",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/the-key-is-the-billing-system.html",
      "startups-and-strategy,site-and-engineering,agents-and-policy",
      "voicedebrief,no-server,client-side,byok,api-keys,billing,openrouter,elevenlabs,speech-to-text,licensing,startups,brief,article",
      "VoiceDebrief turns a voice note into a transcript and a debrief in the browser on the user's own key, with no server by design, and its commercial model was never a billing system: a user pays once, we keep our cut, the rest becomes the credit limit on a key we hand over and the provider meters; OpenRouter's Management API describes that mechanism and its terms forbid reselling access, which is where the project's commit history stops, the same shape as Apple's licence stopping a Mac leased by the minute; the map of every provider's terms and docs shows only OpenRouter and ElevenLabs have keys that carry a ceiling, ElevenLabs has OEM terms and a grants programme, four routes remain, and the brief asks every provider the same sentence: let the credit limit be a property of a key a user can buy."
    ],
    [
      "ways-to-read-one-article",
      "nr:ways-to-read-one-article",
      "Ways to read one article: the finished piece stays at the centre, and the reader picks the way in",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/ways-to-read-one-article.html",
      "news-and-evidence,graphs-and-knowledge,site-and-engineering",
      "newsroom,reading,views,five-readers,infographics,slides,graphs,local-models,personas,mobile,email,brief,article",
      "The views the five readers make, the catalogue, the maps, the arc, the two-minute version and the deck, are good and badly delivered, stacked as folds inside the article; with the pane about the piece the reader picks a mode in one tap and sets a default per persona: headlines, infographic, two minutes, slides, what is new against what they have read, facts, relationships, vocabulary, and, through a model in their own browser, a view at their level; most modes are computed from existing views and the graph, deterministic and explainable, with no round trip to the publisher; and the finished article is kept whole at the centre because it has a beginning, a pacing, a surprise and a voice and is the one artefact that must make sense end to end, which a graph cannot."
    ],
    [
      "firing-journalists",
      "nr:firing-journalists",
      "Firing journalists is the worst thing a publication can do: what a hiring decision says about where a publisher is going, and who to hire instead",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/firing-journalists.html",
      "news-and-evidence,startups-and-strategy",
      "journalism,publishing,hiring,layoffs,strategy,metrics,engagement,provenance,newsroom,roles,article",
      "Against the 2026 job-cut figures, a publisher's hiring decisions are read as the clearest statement of where its centre of gravity is going: firing the people who make original content and hiring for traffic moves the centre to packaging, where measuring engagement makes clickbait by construction because what you measure is what you become; firing the engineers fails the other way, with journalism nothing enables; the case is to keep and tool the journalists, hire the roles that turn a notebook into a graph and the engineers who build what the tooling needs, measure value to the reader, and read any publisher's real strategy from outside with five counts."
    ],
    [
      "the-site-is-an-inbox",
      "nr:the-site-is-an-inbox",
      "The site is an inbox: deciding before reading, marks that build the reader's own graph, and why I could not write this brief until the vault existed",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/the-site-is-an-inbox.html",
      "news-and-evidence,graphs-and-knowledge,vaults-and-method",
      "newsroom,reading,inbox,personas,graphs,feedback-loop,highlights,signals,privacy,vaults,pki,brief,article",
      "With the reader's vault live, the next brief treats the site as an inbox that starts empty: every piece, original or aggregated, waits for a one-tap decision in a fast inbox mode with nothing loaded, and only reading is the slow mode, where marks of six kinds become nodes in a graph the reader owns, connected to the graphs the site publishes for every article; scroll depth, awake time and from-and-to are recorded for the reader and kept as the meter keeps everything, in the browser and the vault; personas each have a key pair so a comment to the newsroom goes down the write-only lane signed with a copy kept; nothing is gamified; and the brief could only be written once the properties of what runs became the constraints to think with."
    ],
    [
      "the-vault-is-live",
      "nr:the-vault-is-live",
      "The vault is live: from a design on 11 October to a reader's vault on newsroom.sgit.ai by two in the morning, and how all of it fits together",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/the-vault-is-live.html",
      "vaults-and-method,site-and-engineering",
      "vaults,passkeys,newsroom,architecture,client-side,sync,local-first,sgit,testing,release,article",
      "Between the morning of 11 October and 02:22 UTC on 12 October the reader's vault went from a design to a running feature on newsroom.sgit.ai: a Vault pane on every page enables the account with a passkey, creates a door at its derived address, a master and a reader vault as real sgit vaults through the vendored vault library, writes each device's raw records as one file in one commit, and brings other devices' activity into the browser so the meter shows one balance; local storage stays primary, the vault holds records never totals, one file per device means no conflict by construction, eight test files run in CI against a mock of the vault server plus a CLI round trip, and a live check on an iPad confirmed it; the properties of what runs are the constraints the next brief is thought with."
    ],
    [
      "two-agents-one-vault",
      "nr:two-agents-one-vault",
      "Two agents, one vault: the Industry Wire is written into an encrypted vault by one agent and drawn by newsroom.sgit.ai, which holds only the read key",
      "2026-10-12",
      "https://newsroom.sgit.ai/articles/two-agents-one-vault.html",
      "vaults-and-method,news-and-evidence,agents-and-policy",
      "vaults,agents,newsroom,industry-wire,read-key,client-side,provenance,personas,workflow,myfeeds,article",
      "A vault is the right interface between two agents that never share a session: the wire desk writes 40 stories, 55 sources, 7 personas and a code-only join into an encrypted sgit vault with a write key the site never holds; newsroom.sgit.ai holds the published read key and about four hundred lines that open the vault in the reader's browser, decrypt it and draw it as text, with 28 checks against the real vault in CI; the site owns the presentation and the vault owns the news, a commit is on the page within two minutes with no release, the server sees ciphertext, the history is the audit trail, and the site's agent read the whole vault before publishing the key and held back the one named reader's lens until that person agreed in writing."
    ],
    [
      "the-article-remains-a-product",
      "nr:the-article-remains-a-product",
      "The article remains a product: Jonathan Palmer's four outputs for a newsroom with no human readers, checked against one that runs",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/the-article-remains-a-product.html",
      "news-and-evidence,graphs-and-knowledge,agents-and-policy",
      "newsroom,future-of-news,structured-intelligence,provenance,graphs,versions,agents,llms-txt,vaults,world-news-day,economics,response,article",
      "Palmer's four outputs, structured intelligence, continuous intelligence, provenance and verification, and embedded distribution, are what a newsroom built on the story-as-graph thesis already produces: graphs as JSON beside all 74 articles, version histories read from git, a hashed corpus in a vault with a published read key, and an llms.txt and markdown twin for agents; what is designed and not running, selling the graph, propagating corrections and paying the fact creator, is said plainly; and where the two differ is the word proprietary, because provenance only someone else can check is worth charging for."
    ],
    [
      "explorer-to-villager-newsroom",
      "nr:explorer-to-villager-newsroom",
      "Go slow to go fast: moving the newsroom from an explorer project to a villager one, with the brief to do it",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/explorer-to-villager-newsroom.html",
      "site-and-engineering,startups-and-strategy",
      "engineering,explorer,villager,refactoring,testing,ci,security-review,code-review,newsroom,sg-meter,github-issues,brief,article",
      "The newsroom's reader features are live and taking money, but bundled with the content in one repository; measured at v0.8.0, about 2,700 lines of live reader code have 433 lines of tests the one workflow never runs, two upstreams for the same code (sgit.ai and the newsroom, with diverging meters) and no Content-Security-Policy. To keep shipping fast without breaking anything, the reader application moves to its own repository, which both sites consume, as a pure refactoring, proved by the live site serving the same bytes, then goes through the explorer-to-villager phases: map, test with bugs as passing tests and no mocks, review with every finding a GitHub issue, refactor until no adversarial change slips through, and only then new features. A pipeline under two minutes per commit runs Node and three browser engines, with pre-production on a Stripe test link, synthetic readers and a live check at release."
    ],
    [
      "the-passkey-opens-the-vault",
      "nr:the-passkey-opens-the-vault",
      "The passkey opens the vault: a door per path, a master vault for every user, and no server of ours",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/the-passkey-opens-the-vault.html",
      "vaults-and-method,site-and-engineering",
      "passkeys,webauthn,prf,vaults,key-management,pki,client-side,newsroom,architecture,brief,secrets,lab,article",
      "An end-to-end, client-side account with no sign-in to the site: a passkey with the WebAuthn PRF extension gives 32 bytes only the user's authenticator can produce for the site, and because sgit vault keys are made in the client the browser computes a door vault's key and address from them, so nothing is stored or found first; the door holds the master vault's key and a record of its provider, path, credential id, device and date, and the master holds the keys of the vaults below, received read keys, the access key, key pairs generated at random and never derived from the passkey, and the list of its doors. A test on newsroom.sgit.ai on 11 and 12 October 2026 held for the platform path, the derivation and a door on a real server, and showed that the same passkey reached by a phone over QR gives different bytes, so the claim is narrowed to the same kind of path and the unit of the design is a door per path: added in the first week while unlocked with no key typed, bootstrapped by the recovery code where no device has both paths, with an unknown address meaning no door here yet and a door retired by deleting its file, never by destroying its derived address."
    ],
    [
      "i-miss-it",
      "nr:i-miss-it",
      "I miss it, so it is working: the people who decide the roadmap have to use the product, and the vault the newsroom now needs",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/i-miss-it.html",
      "startups-and-strategy,site-and-engineering",
      "product,dogfooding,startups,feedback-loop,execution-speed,vaults,sync,pki,passkeys,newsroom,personas,brief,article",
      "The people who decide a roadmap have to use the product: they give the best feedback and direction, and when they do not use it, something is missing. Dinis failed that test with earlier startups and the vaults now pass it, getting better because they are used. With agents able to build almost anything, direction from usage is the scarce part, and the speed of the loop from comment to article to feature is the moat. Reading the newsroom on four devices that do not agree is the signal for the next feature: the reader's own encrypted vault, with an access key, a key pair, a log per device merged by union, a way to add a device, and a one-way line to the agent that edits the site."
    ],
    [
      "brief-the-agent-too",
      "nr:brief-the-agent-too",
      "Brief the agent too: the people I talk to work with agents, so what I send has to reach both",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/brief-the-agent-too.html",
      "vaults-and-method,news-and-evidence",
      "agents,briefs,reading-list,personas,memory,compression,tokens,onboarding,newsroom,mentoring,agent-to-agent,prompt-injection,article",
      "Readers increasingly read through an agent. A founder gave an article on threat tiers to their agent, which saw the team defending against states while backups were undone and reset the target to T3. So content should travel by several routes: to the person, to their agent as a plain-text brief with the articles by number, reasons, graph slices and rules for critical reading, back from the agent to the person as a translation, and later agent to agent through an encrypted lane. Articles are compression we control, which keeps workflows affordable; an agent with an account and memory should read only the delta, which is what it would pay for. Packs must be written safely: data not instructions, nothing hidden, nothing that acts for us, disagreement invited."
    ],
    [
      "a-reading-list-in-a-link",
      "nr:a-reading-list-in-a-link",
      "From me to you, in a link: a personal reading list, two briefs to build and use it, and the loop that makes it a day's work",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/a-reading-list-in-a-link.html",
      "site-and-engineering,startups-and-strategy",
      "newsroom,personas,curation,reading-list,briefs,crm,outreach,workflow,feedback-loop,agents,product,article",
      "A personal reading list in a URL fragment (from, to, article numbers to read, to look at, and to follow) that opens on the newsroom as a persona the reader can keep or remove, with a stable public number for every article and a page to make a list for anyone. Two briefs: the newsroom builds it with validation, tests and no tracking; the RiskMandate agent, which keeps the contacts, chooses the person with Dinis, researches public sources, picks articles with reasons, tests the link, drafts the email, and Dinis sends it. The core is the loop: editorial, product, engineering, sales and the founder connected by written briefs, small contracts, owned repositories and tests, closing in about a day, as it did this week, with costs that are written down: drift, budgets and one person as the router."
    ],
    [
      "a-model-in-the-tab",
      "nr:a-model-in-the-tab",
      "A model in the tab: Chrome now ships an LLM to capable desktops, and it is the hardest piece of the client-side puzzle",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/a-model-in-the-tab.html",
      "site-and-engineering,graphs-and-knowledge",
      "client-side,gemini-nano,prompt-api,chrome,built-in-ai,on-device,privacy,sovereignty,interoperability,webmcp,newsroom,personas,article",
      "Since Chrome 148 any page can prompt Gemini Nano on the reader's device, which supplies the hardest missing piece of the client-side pattern; the rough first run is mostly fixable by the page, the scarcity of examples has real causes, the standards objections are serious, and the way to build on it without being captured is to keep the meaning in anchored graphs and use the model as a checkable function."
    ],
    [
      "no-server-by-design",
      "nr:no-server-by-design",
      "No server, by design: first I retired the database, now I am retiring the back end",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/no-server-by-design.html",
      "site-and-engineering,vaults-and-method",
      "architecture,client-side,serverless,no-database,lets,file-system,vaults,append-lanes,privacy,sovereignty,security,newsroom,sg-meter,wardley,article",
      "After replacing the database with the file system (LETS), the next step is removing the back end: encrypted vaults make the server blind, so the application moves to the browser, leaving only commodity products on the server side; the security gaps this leaves are written down and accepted because they do not match the threat, and the result scales without its complexity scaling and ships as files."
    ],
    [
      "the-agent-is-the-reader",
      "nr:the-agent-is-the-reader",
      "The agent is the reader: why agents will pay for graphs, personas and signed claims, because it is cheaper than not paying",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/the-agent-is-the-reader.html",
      "news-and-evidence,startups-and-strategy",
      "newsroom,economics,agents,micropayments,fractal-semantic-graphs,compression,personas,tokens,provenance,signed-claims,librarian,cartographer,article",
      "When agents are the readers, a publisher that serves curated meaning, an anchored graph fitted to the agent's persona and objective, and claims signed with an expiry saves the agent more tokens than it charges, so the agent pays not to pay; measured on five articles the graph answers in 90% fewer tokens, and on 70 articles compression with scale appears only where a Cartographer curates, which is the work worth paying for."
    ],
    [
      "where-the-platform-draws-the-line",
      "nr:where-the-platform-draws-the-line",
      "Where the platform draws the line: Microsoft Execution Containers, the shared responsibility model for agents, and the business logic above it",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/where-the-platform-draws-the-line.html",
      "agents-and-policy",
      "riskmandate,agent-behaviour-policy,microsoft,mxc,execution-containers,agent-control-specification,containment,shared-responsibility,business-logic,barriers,briefing,article",
      "Microsoft Execution Containers and the Agent Control Specification confirm the direction of travel, policy outside the agent, deny by default, observe before enforcing, but like the cloud's shared responsibility model they draw the line where the platform can see, at paths, numeric addresses and tool calls, so the business logic above it, what a label means, whose voice, how many, which customer, and who accepts what is left, has to come from the deployer's behaviour policy, which is what turns the platform's hooks into boundaries."
    ],
    [
      "the-same-argument-in-our-words",
      "nr:the-same-argument-in-our-words",
      "The same argument, in our words: Satya Nadella on models as insider risks, translated into the language of RiskMandate",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/the-same-argument-in-our-words.html",
      "agents-and-policy",
      "riskmandate,agent-behaviour-policy,insider-threat,mandate,grant,barriers,enforcer-test,risk-acceptance,software-was-the-law,philosophy,response,article",
      "Nadella's argument and ours are the same argument in two vocabularies: his lost code path is software was the law, his insider is the ultimate insider, his separation of supply from authority is the grant is not the mandate, his controls outside the model are the enforcer test, and his last line is our brakes; the languages differ in unit, a model against one agent in one deployment, and in frame, observation against authority accepted by a named person for a stated interval."
    ],
    [
      "authority-outside-the-model",
      "nr:authority-outside-the-model",
      "Authority outside the model: an answer to Satya Nadella on models as insider risks, principle by principle, with what we run, what we ship and what we plan",
      "2026-10-11",
      "https://newsroom.sgit.ai/articles/authority-outside-the-model.html",
      "agents-and-policy",
      "riskmandate,agent-behaviour-policy,insider-threat,containment,observability,mandate,barriers,second-reader,connector-twin,footprint,incident-disclosure,licence-to-operate,synthetic-users,response,article",
      "Treating models as insider risks is right, and each of the seven principles maps to something this network runs, publishes or plans, but the principles work only with the column they leave thin: authority written down before the agent runs, a mandate compared with the footprint, risk belonging to one deployment rather than a model, a named owner and expiry for what is left, recoverability beyond the brake, and decisions made in advance as envelopes."
    ],
    [
      "how-to-run-synthetic-users",
      "nr:how-to-run-synthetic-users",
      "How to run synthetic users on your own site: five people who do not exist, a browser, and an afternoon",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/how-to-run-synthetic-users.html",
      "vaults-and-method,site-and-engineering",
      "synthetic-users,ux,testing,personas,agents,claude,playwright,vaults,newsroom,how-to,article",
      "Synthetic users are invented people, each a record with a situation, a question, what they know and do not, what they decide on and what makes them leave, played by a model that drives a real browser one screenshot at a time instead of reading the page's code, so it can get lost as a person would. Claude Sonnet is good enough to play each user; the study lives in a GitHub repository or a vault shared by a link. Choose five to ten users of about five types, write profiles and expected journeys first, run the observe-think-act loop, interview and rate, order findings by cost, check every surprising finding against the browser's record, publish and run again. On newsroom.sgit.ai five readers produced 31 findings, six blocking; two were caused by the helper, and one became a measured bug: 51 of 527 figures missing."
    ],
    [
      "a-link-to-a-persona",
      "nr:a-link-to-a-persona",
      "A link to a persona: send people to the newsroom as someone, not as nobody",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/a-link-to-a-persona.html",
      "startups-and-strategy,vaults-and-method",
      "newsroom,personas,personalisation,vaults,local-first,privacy,onboarding,agents,brief,article",
      "A link should land a reader as someone rather than as nobody: personas move from the site's code into an encrypted vault kept by the newsroom's agents and opened in the browser with a published read key, reached through a page of their own, a URL fragment that never reaches a server, or a private vault for one person, and the reader can follow the persona as it is updated or fork it as their own, all client side."
    ],
    [
      "pay-after-you-read",
      "nr:pay-after-you-read",
      "Pay after you read: how a reading meter became a working business model in one afternoon, one release at a time",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/pay-after-you-read.html",
      "news-and-evidence,startups-and-strategy",
      "newsroom,reading-meter,micropayments,pricing,personalisation,personas,wardley-maps,innovation,agents,article",
      "On 10 October the reading meter became a working pay-on-demand publication in seven releases in 3 hours 40 minutes: pay for the share read from a balance that can go below zero, rate usefulness after reading to set the price from free to double, personas and newsrooms kept in the reader's browser, and consented, encrypted sharing for a designed front page. Paying after the value inverts Leanpub's pay-before model and is supported, with caveats, by pay-what-you-want evidence; Blendle's refunds show ratings measure quality. Each feature grew from using the last, which a Wardley map shows: new components on the left, built on commodities already there. It was fast because of non-functional work done before. Every decision started from not ripping off the reader. What is missing is traffic."
    ],
    [
      "where-the-vault-keys-live",
      "nr:where-the-vault-keys-live",
      "Where the vault keys live: key management at sgit-ai v0.20.0, and what comes next",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/where-the-vault-keys-live.html",
      "vaults-and-method,agents-and-policy",
      "sgit,vault-keys,key-management,secrets,append-lanes,pki,passkeys,password-managers,agents,article",
      "Vaults are encrypted in the client, so the server never holds a key and the key is the whole question. Today vault keys live in a password manager and in a registry vault run by an isolated session; agents that create vaults seal the key to the registry's public key and send it on a write-only append lane, so it never enters a chat. The weakest step is giving a key to another agent, because agent platforms have no per-session secrets. Append lanes carry keys, logs, observations and messages; small communication vaults keep each key's blast radius small and multiply the keys. Next: password manager integrations, secrets.sgit.ai's passkey-unlocked keyring, PKI, and out-of-band decryption."
    ],
    [
      "the-infographic-bake-off",
      "nr:the-infographic-bake-off",
      "The infographic bake-off: which image model for which job, judged blind on 10 October 2026",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/the-infographic-bake-off.html",
      "site-and-engineering,news-and-evidence",
      "infographics,image-models,openrouter,gemini,gpt-image,evaluation,cost,newsroom,designer,accountant,data-scientist,article",
      "To make infographics part of the newsroom's workflow, every image-output model on OpenRouter on 10 October 2026 was given the same briefs from one article, from 18 words to 2,800, plus an edit, a component, a brand slide and a deck, over five rounds with blind judging and recorded costs; the answer is to choose the model by the job, not the budget: Nano Banana 2.1 for concept slides, charts and decks, Gemini 3 Pro Image when speed matters, GPT-5.4 Image 2 for documents, edits, components and brand slides, with prompt rules learned from the faults and a style reference for every deck."
    ],
    [
      "pay-to-keep-your-persona",
      "nr:pay-to-keep-your-persona",
      "Pay to keep your persona: readers should pay because it helps them, not because they feel they should",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/pay-to-keep-your-persona.html",
      "startups-and-strategy,news-and-evidence",
      "newsroom,personalisation,personas,value-proposition,micropayments,local-first,privacy,focus,article",
      "Paying for content because you have to, feel you should, or think it right works to a degree but does not scale and makes readers feel the wrong thing; what readers would pay for is what helps them, time, context, focus and good decisions, carried by a persona built from their reading, named, drawn as a graph, curated, and multiplied for focus, and the paid product is that persona following them between devices and agents with the privacy intact."
    ],
    [
      "who-will-game-the-reading-meter",
      "nr:who-will-game-the-reading-meter",
      "Who will game the reading meter? Eight kinds of reader, nine ways to cheat, and the risks that will actually happen",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/who-will-game-the-reading-meter.html",
      "startups-and-strategy,agents-and-policy",
      "security,threat-modelling,micropayments,local-first,privacy,usability,users,agents,article",
      "A reading meter kept entirely in the browser can be cheated by anyone, but the people who could cheat are readers, not attackers; sized on a ladder from invisible engineers (about 0.6% of internet users are developers) to people who cannot use a computer (26% of OECD adults), with agents moving between steps, the cheating costs the site nothing it would have had, while the risks that will actually happen are confusion, a shared computer exposing history and card fraud, so the design is aimed at the readers who will pay and at not frightening the rest."
    ],
    [
      "going-live-with-the-reading-meter",
      "nr:going-live-with-the-reading-meter",
      "Going live with the reading meter: pay for what you read, go below zero if you like, and the numbers we will check in eight weeks",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/going-live-with-the-reading-meter.html",
      "startups-and-strategy,news-and-evidence",
      "newsroom,micropayments,pricing,personalisation,local-first,business-case,hypotheses,stripe,article",
      "The reading meter goes live as an honesty box rather than a paywall: pages are charged by scroll depth, the balance may go below zero forever, there is no ask and no banner, readers can decline to pay with a reason, and topping up is one £5 Stripe payment credited by an unverified return page; the business case is evidence, not revenue, and five hypotheses with thresholds and a review date are written down before the first payment."
    ],
    [
      "open-source-is-not-free",
      "nr:open-source-is-not-free",
      "Open source is not free: who pays to keep the long tail working?",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/open-source-is-not-free.html",
      "startups-and-strategy,vaults-and-method",
      "open-source,economics,funding,long-tail,support,micropayments,simulation,data-science,homebrew,docker,macos,article",
      "Open source is not free: somebody pays, at least in time. The long tail of old versions and old machines is measurable (6.99% of Homebrew macOS install events in 30 days, 16.66% in a year, are on versions older than the newest three) and projects drop it because it has no revenue stream, while vendors such as Oracle and Red Hat turn the same need into large, high-margin support businesses. Pence payments only work aggregated, because a 20p card fee dominates below about five pounds. Simulated over 10,000 draws for one project at Homebrew's scale, micro-paid supported builds cover the long-tail cost in half the draws at about 1.80 pounds a machine a year with typical conversion but never in nine in ten, donations in about a quarter, and customised builds for companies in 94%: the reliable money is companies paying for a build below their own cost of keeping the tool working, sold as value, not as a donation or a weaker free tier."
    ],
    [
      "one-article-five-readers",
      "nr:one-article-five-readers",
      "One article, five readers: a librarian, a cartographer, a historian, an explainer and a storyteller read the same piece",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/one-article-five-readers.html",
      "news-and-evidence,graphs-and-knowledge",
      "newsroom,journalism,liquid-content,fractal-semantic-graphs,ontology,agents,librarian,cartographer,historian,multi-view,article",
      "Long, evidence-rich articles are hard to consume, and the answer is not shorter articles but extraction after the writing: the prose stays the source because writing is how the argument is found, and five agent readers with defined roles then produce other views of it, two building the platform (a Librarian that catalogues every item with its verbatim anchor, a Cartographer that builds an ontology at article, topic and site altitude and draws the maps) and three writing for a person (a Historian on what the article added and where it sits, an Explainer in two minutes, a Storyteller as a deck); run on five articles it catalogued 662 items, flagged 32 problems in published text, merged ideas named differently across articles, and measured each article's contribution as new concepts falling 17, 11, 5, 11, 1."
    ],
    [
      "a-meter-in-the-browser",
      "nr:a-meter-in-the-browser",
      "A meter in the browser: a penny a page, a history you keep, and a site that picks for you",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/a-meter-in-the-browser.html",
      "startups-and-strategy,news-and-evidence",
      "newsroom,micropayments,personalisation,local-first,privacy,pricing,early-access,article",
      "A reading meter that lives only in the browser, a few pence a page from five pounds of starting credit with a simulated top-up and no blocking, tests whether people would pay a little for reading when what they get back is a site that personalises itself from a history nobody else holds; escaping the meter is easy, but escaping it also discards the history, which is the point of the experiment."
    ],
    [
      "re-anchoring-agent-behaviour-policies",
      "nr:re-anchoring-agent-behaviour-policies",
      "Re-anchoring: keeping an agent's rules alive through every summary, and a canary that shows when they are not",
      "2026-10-10",
      "https://newsroom.sgit.ai/articles/re-anchoring-agent-behaviour-policies.html",
      "agents-and-policy,site-and-engineering",
      "agents,claude-code,hooks,agent-behaviour-policy,compaction,context,re-anchoring,guardrails,governance,article",
      "Instructions given only in conversation can be lost when a long session is summarised, and this site's own session was summarised eighteen times in a month, keeping about 1.7% each time; re-anchoring keeps the agent's rules in an Agent Behaviour Policy file and has a SessionStart hook on the compact matcher print it back after every summary, so rules are restored rather than remembered, while a canary status report computed from the transcript, checked by a Stop hook and expected by the person, shows the policy is still being read; the same file serves as memory, checker rulebook and enforcement list, and the best recipe for each rule changes with every platform release, which is the work worth curating."
    ],
    [
      "a-second-reader-the-agent-cannot-skip",
      "nr:a-second-reader-the-agent-cannot-skip",
      "A second reader the agent cannot skip: how two wrong emails became a gate on every draft",
      "2026-10-09",
      "https://newsroom.sgit.ai/articles/a-second-reader-the-agent-cannot-skip.html",
      "agents-and-policy,site-and-engineering",
      "agents,claude-code,hooks,guardrails,llm-as-a-judge,maker-checker,agent-behaviour-policy,email,governance,article",
      "A rule held by the agent it governs is an expectation that can be forgotten, missed or argued around, so the inbox agent's fix after two emails went out in the founder's voice puts a gate on the draft tool itself: deterministic code checks first, then a fresh model call with only the rules, the sources and the draft, a different model from the author, failing closed on anything but an explicit PASS and logging every verdict; its independence depends on where it is installed, a setting while the agent can change it and a boundary only under managed placement, and the pattern is the old maker-checker control applied to any consequential tool call."
    ],
    [
      "the-ai-governance-stack-as-a-graph",
      "nr:the-ai-governance-stack-as-a-graph",
      "The AI governance stack, as a graph: an answer to Hari Kota, built",
      "2026-10-09",
      "https://newsroom.sgit.ai/articles/the-ai-governance-stack-as-a-graph.html",
      "graphs-and-knowledge,agents-and-policy",
      "ai-governance,fractal-semantic-graphs,graphs,eu-ai-act,iso-42001,nist-ai-rmf,risk,inventory,competence,vaults,article",
      "Keep Hari Kota's ten-layer stack exactly as posted as the view and make a fractal semantic graph the model: the table already contains ten edges across layers where rows share an instrument, role or jurisdiction; each layer is its own world with its own ontology, sharing only a grammar of verbs with inverses; the edges between worlds come from the provisions themselves, such as EU AI Act Article 26(2) and NIST AI RMF GOVERN 1.6; changes over time are dated edges rather than rewrites; and a gap is a missing edge, so the do-this-today test and the gaps in layers 3, 5 and 10 become queries anyone can run."
    ],
    [
      "rfc-0001-public-key-cryptography-for-sgit",
      "nr:rfc-0001-public-key-cryptography-for-sgit",
      "RFC 0001: two ways to add public-key cryptography to sgit, and the questions we want you to answer",
      "2026-10-09",
      "https://newsroom.sgit.ai/articles/rfc-0001-public-key-cryptography-for-sgit.html",
      "vaults-and-method,site-and-engineering",
      "rfc,sgit,pki,public-key-cryptography,encryption,signatures,ed25519,x25519,hpke,age,sealed-files,key-management,threat-model,request-for-comments,article",
      "An sgit vault's read key is symmetric, so anyone who holds it and can write to the store can author history clients accept; turning the read key into a public key does not fix that, but two key pairs do, giving a native PKI vault mode where readers cannot write, hosts cannot forge and write-only depositors become possible, while an optional sealed-files layer encrypts chosen files to named people's public keys with the private key held in a file, an ssh key, hardware or a remote service that logs and approves each decryption; both share a first step, a writer signature on today's vaults, and nothing is built until reviewers have answered fourteen questions."
    ],
    [
      "a-mac-of-the-agents-own",
      "nr:a-mac-of-the-agents-own",
      "A Mac of the agent's own: a business plan for agent desktops, what Apple's licence allows, and three behaviour policies",
      "2026-10-09",
      "https://newsroom.sgit.ai/articles/a-mac-of-the-agents-own.html",
      "startups-and-strategy,agents-and-policy,vaults-and-method",
      "agents,agent-desktops,macos,business-plan,agent-behaviour-policy,licensing,vaults,pki,sgit,wardley-maps,isolation,cowork,computer-use,article",
      "After a mailbox, a code-host account and a Claude account, an agent's next dedicated resource is a desktop, and it should be a Mac; Apple's licence rules out a pool rented by the minute, so the business is a dedicated Mac per customer, run for them, metered by the minute, built clean per run from encrypted vaults with keys by PKI and erased at the end, plus developer agents, software for your own Mac and a request to Apple for terms; the behaviour policies show why: the same agent's unbounded excess falls from 23 rows to 11 to 3."
    ],
    [
      "riskmandate-ten-questions",
      "nr:riskmandate-ten-questions",
      "Ten hard questions for RiskMandate, answered: the mandate, the reach, the gap, and what we are deliberately not",
      "2026-10-09",
      "https://newsroom.sgit.ai/articles/riskmandate-ten-questions.html",
      "agents-and-policy,startups-and-strategy",
      "riskmandate,agent-behaviour-policy,mandate,grant,delta,barriers,insurance,liability,enforcement,supply-chain,pki,digital-twins,metrics,positioning,interview,article",
      "Every answer rests on four objects for one agent in one deployment, mandate, grant, delta and barriers, and a strict division of roles: RiskMandate authors the policy and pre-commits the verdict, the customer's controls enforce it, a named executive owns the residual risk, and logs calibrate the grant; so RiskMandate is not inline, not an insurer and not one button, the value lands before the action, new capability can only land as excess, and the headline metric is the reduction in risk the business has to accept."
    ],
    [
      "how-i-work-with-claude",
      "nr:how-i-work-with-claude",
      "How I work with Claude: one session per topic, agents with names, and memory you curate",
      "2026-10-09",
      "https://newsroom.sgit.ai/articles/how-i-work-with-claude.html",
      "agents-and-policy,vaults-and-method",
      "claude,workflow,sessions,agents,memory,context-management,onboarding,prompts,vaults,agent-behaviour-policy,cowork,guide,article",
      "Working well with Claude comes from a few habits: one session per major or recurring topic and no mixed threads, a naming pattern with @ names for agents, an agent defined as a focused session with a role.md, memory treated as what the session reads and curated as websites, graphs and vaults wired together and accessible by default, vaults for sharing by key instead of broad permissions, review as the quality step that fixes the brief, and an Agent Behaviour Policy before every new connector."
    ],
    [
      "the-waiting-room-knew-first",
      "nr:the-waiting-room-knew-first",
      "The waiting room knew first: a live local story, and the gap where local information used to be",
      "2026-10-08",
      "https://newsroom.sgit.ai/articles/the-waiting-room-knew-first.html",
      "news-and-evidence,vaults-and-method",
      "local-news,live-evidence,nhs,hospitals,public-information,it-outage,evidence-levels,vaults,agents,story-vault,news-deserts,article",
      "On 8 October 2026 staff at two sites of one London trust told patients the IT systems were down, a five-hour wait was announced and a doctor called it national, yet nothing a local person could check before leaving home showed any of it; the story is not the outage or the hospital but the gap in local information: some UK nations publish live emergency waits, England has no national equivalent, local news has thinned and social feeds have fragmented, so the evidence is kept in a vault by level, with sources re-checked, an enquiry to the trust and the answer logged, as the first live counterpart to the fictional bridge story."
    ],
    [
      "encrypted-memory-for-isolated-agents",
      "nr:encrypted-memory-for-isolated-agents",
      "Encrypted memory for agents that run somewhere else: sgit deployment patterns, from a Mac mini to Kubernetes",
      "2026-10-08",
      "https://newsroom.sgit.ai/articles/encrypted-memory-for-isolated-agents.html",
      "vaults-and-method,agents-and-policy,site-and-engineering",
      "agents,agentic-memory,self-hosting,deployment,docker,docker-compose,kubernetes,mac-mini,aws,cloudfront,fargate,lambda,zero-knowledge,encryption,partial-clones,sgit,isolation,article",
      "Agents are safer in isolated places, but isolation takes away the shared drive and ephemeral compute takes away the disk, so their memory has to live somewhere the owner controls and a breach does not expose; a self-hosted SG/Send container with an access token and storage in a folder or bucket, with sgit in each agent, gives them that: scoped clones make each run fast, the server holds only ciphertext, any clone can restore any server, and the trust sits with the agents and their keys, so a scoped clone is speed, not a boundary, and one vault per trust boundary is the rule."
    ],
    [
      "knowing-when-to-stop",
      "nr:knowing-when-to-stop",
      "Knowing when to stop: what experience gives people, and what we have to design into agents",
      "2026-10-08",
      "https://newsroom.sgit.ai/articles/knowing-when-to-stop.html",
      "agents-and-policy,startups-and-strategy",
      "agents,knowing-when-to-stop,diminishing-returns,seniority,security,threat-modelling,kpis,goodhart,wardley-maps,constraints,fire,shipping,agent-behaviour-policy,article",
      "Knowing when to stop is the hardest call in most work: for people it comes from perspective, knowing the attackers, the phase of the business, the bottleneck and what good enough looks like, which is much of what seniority is; good teams without it defend against movie plots or hit every KPI without moving the business; agents have the same problem by design because their range is the feature, so they keep going; the answer for both is constraints that carry perspective, direction, a mandate, memory with the bigger picture, graphs that narrow the scope, a stop named in advance, one kind of work per step, a reviewer with agency, and shipping often enough to see whether it mattered."
    ],
    [
      "agency-is-not-a-yes",
      "nr:agency-is-not-a-yes",
      "Agency is not a yes: a scale for human and agent decisions, from rubber stamp to the reviewer who fixes the source",
      "2026-10-08",
      "https://newsroom.sgit.ai/articles/agency-is-not-a-yes.html",
      "agents-and-policy,graphs-and-knowledge",
      "agency,accountability,human-in-the-loop,decision-making,agents,agent-behaviour-policy,provenance,fractal-semantic-graphs,eu-ai-act,maturity-model,vaults,article",
      "Agency for anyone asked to decide, a person or an agent, needs options beyond yes, context in their own terms, the ability to look behind what they are shown, time or tokens, symmetric incentives, escalation and authority over the source; the weakest of these caps the decision on a seven-level scale; below level 3 the decider carries liability rather than agency and accountability belongs to whoever designed the decision point and up their chain; above it the review is a QA step that validates everything upstream and fixes the source, and at the top proven cases become rules that run as code under a behaviour policy, which is also what the EU AI Act's oversight requirements already ask for."
    ],
    [
      "hope-or-enforcement",
      "nr:hope-or-enforcement",
      "Hope or enforcement: one customer service agent, three designs, and who keeps each promise",
      "2026-10-08",
      "https://newsroom.sgit.ai/articles/hope-or-enforcement.html",
      "agents-and-policy,vaults-and-method",
      "agents,agent-behaviour-policy,customer-service,multi-agent,orchestration,harness,blast-radius,prompt-injection,tokens,riskmandate,simulation,vaults,article",
      "Every design of an agent can have a rule for every behaviour; what differs is who enforces each rule and how much the agent can reach, so refactoring a customer service agent from one model with broad connectors, to a harness of business tools, to a team of narrow agents behind a deterministic identity gateway turns most of its policy from expectations the model keeps into boundaries it cannot reach, shrinks the policy because it no longer has to forbid what is impossible, cuts tokens, and leaves a short map of what still rests on hope and which boundary to build next, read after every run and shown to the client as a record rather than a score."
    ],
    [
      "who-are-you-protecting-against",
      "nr:who-are-you-protecting-against",
      "Who are you protecting against? Draw the security line where the attacker is, not above it",
      "2026-10-08",
      "https://newsroom.sgit.ai/articles/who-are-you-protecting-against.html",
      "agents-and-policy,startups-and-strategy",
      "security,threat-modelling,threat-agents,attack-trees,mitre-attack,nist,ncsc,startups,founders,air-gap,cloud,vaults,article",
      "Security decisions start with who the threat agent is, what the attack vector is and how sophisticated they are; read against a ladder from your own mistakes to elevated threats, grounded in NIST SP 800-30, the NCSC and MITRE ATT&CK, most startups should plan up to organised crime for money, do the basics consistently because they beat most attackers, sell something better than the customer's baseline rather than beyond their needs, prefer isolation and visibility over disconnection, and write down what they accept above their ceiling, because drawing the line too high slows everything, can create new holes and can stop the technology being used."
    ],
    [
      "every-mistake-added-a-rule",
      "nr:every-mistake-added-a-rule",
      "Every mistake added a rule: complexity, agents, and the way back to shipping",
      "2026-10-08",
      "https://newsroom.sgit.ai/articles/every-mistake-added-a-rule.html",
      "agents-and-policy,site-and-engineering",
      "agents,complexity,wardley-maps,engineering-practice,context-window,compaction,memory,shipping,founders,nfrs,testing,cowork,chatgpt,article",
      "Founders who now work as engineers with agents often meet complexity not in their code but in the process around it, and answer each mistake with a rule, which grows the prompt, brings compaction sooner and breeds new mistakes; on a Wardley map that complexity is a position, custom-built process sitting where commodities already exist, and the way out is to make each piece a small, tested, shipped component that moves right and no longer needs a rule, with short sessions, memory as versioned files, security matched to assets and attack vectors, rules only for incidents and machines for enforcement, and shipping as the discipline that keeps blast radius and over-engineering under control."
    ],
    [
      "liquid-content-needs-water",
      "nr:liquid-content-needs-water",
      "Liquid content needs water: liquefy the journalist's notebook, not the finished product",
      "2026-10-07",
      "https://newsroom.sgit.ai/articles/liquid-content-needs-water.html",
      "news-and-evidence,graphs-and-knowledge",
      "journalism,liquid-content,story-vault,fractal-semantic-graphs,newsroom,personalisation,micropayments,agents,ft-strategies,article",
      "Liquid content, journalism built from datafied components, is right in its instincts, but what makes content liquid is the reporting inside it, so the way in is not a reinvention of newsroom norms but putting the experienced journalist and their existing workflow at the centre, liquefying their notebook into a story graph as they work, keeping the writing theirs because writing is how the story is found, bringing in experts paid for time and credibility, personalising by intersecting the reader's graph with the story's, keeping editorial direction rather than readers' tastes as the brief, and getting paid per use down the graph rather than only through advertising, subscriptions and licensing."
    ],
    [
      "the-bridge-followed-to-the-end",
      "nr:the-bridge-followed-to-the-end",
      "The bridge, followed to the end: what one local story is worth when it is kept as a graph",
      "2026-10-07",
      "https://newsroom.sgit.ai/articles/the-bridge-followed-to-the-end.html",
      "news-and-evidence,graphs-and-knowledge",
      "journalism,local-news,reader-skills,liquid-utility,story-vault,fractal-semantic-graphs,micropayments,agents,simulation,article",
      "Played out as a simulation on a story vault, one local bridge closure shows the value of journalism as getting a disputed fact right and keeping it right: newsworthy on five days but needed on fifty-seven, used through Reader Skills by readers whose own graphs meet the story graph at shared anchors and who decide better with it, bought by institutions and agents who need the same claims, cheaper for agents than searching, and paid for in small amounts that walk back down the claims to the reporter, the paper and the residents whose evidence the graph names."
    ],
    [
      "story-vault-meets-reader-skills",
      "nr:story-vault-meets-reader-skills",
      "Story vault underneath, Reader Skills on top: why local journalism has the most to gain",
      "2026-10-07",
      "https://newsroom.sgit.ai/articles/story-vault-meets-reader-skills.html",
      "news-and-evidence,graphs-and-knowledge",
      "journalism,local-news,liquid-utility,reader-skills,story-vault,fractal-semantic-graphs,provenance,micropayments,verification,article",
      "Markus Franz's Liquid Utility proposes Reader Skills that help people understand, update, relate, compare, follow and act on journalism; the story vault keeps reporting as a versioned graph of claims tied to hashed evidence; put together, each skill is an operation on the graph and each safeguard he asks for is a property the graph already has, his bridge closure becomes a graph in which one supersede edge is the update, the alert and the correction, and his emphasis on locality supplies what the vault's economics had not mapped: local contributors and local journalists sit where trust is strongest, stories travel up to national and international outlets, and if every use pays back down the chain of claims it rests on, small payments from many people fund the reporting nearest to them."
    ],
    [
      "the-behaviour-policy-is-the-business-logic",
      "nr:the-behaviour-policy-is-the-business-logic",
      "Zoom into an agent's behaviour policy and you find the business logic",
      "2026-10-07",
      "https://newsroom.sgit.ai/articles/the-behaviour-policy-is-the-business-logic.html",
      "agents-and-policy",
      "agents,agent-behaviour-policy,business-logic,fractal-semantic-graphs,controls,risk-acceptance,rbac,skills,integrations,article",
      "The first rules written for an agent are mechanical and vendors enforce them well, but every layer above them is the company's own business logic: how a function is done, the steps a process goes through, who a client is right now, and what the company is for; much of that logic was never written down because software enforced it by omission, and agents working through APIs bypass that; a behaviour policy built in layers, each refining the one below with its own owner, can hold it, can count how much of it is backed by a control, by an accepted risk or by hope, and gives vendors a written rule to enforce and a number to move, which is what lets the business hand work to agents and scale."
    ],
    [
      "an-agent-desktop-by-the-minute",
      "nr:an-agent-desktop-by-the-minute",
      "A locked-down desktop for an agent, by the minute, is still hard to rent",
      "2026-10-07",
      "https://newsroom.sgit.ai/articles/an-agent-desktop-by-the-minute.html",
      "agents-and-policy,startups-and-strategy",
      "agents,agent-desktops,computer-use,sandboxes,isolation,egress,secrets,macos,windows-365,startup-programmes,agent-behaviour-policy,article",
      "A desktop safe to hand to an agent needs isolation, an egress allowlist, secrets kept outside, its own identity, a live view, a record, a clean reset, billing that follows the work and the right operating system; in October 2026 each exists somewhere and no product we read has all nine; Linux sandboxes come closest, with key controls in beta, Windows agent desktops arrived in June, and macOS cannot be leased for less than 24 hours under Apple's licence, which also limits leased Macs to developer services; because prompt injection is not solved the desktop has to be the barrier, so the proposal is one mandate per task, a fresh desktop per task by operating system, and a vault outside every desktop, paid for at first by startup programmes listed with how to apply."
    ],
    [
      "ai-baseline-control-framework",
      "nr:ai-baseline-control-framework",
      "An open AI governance framework, and what its licence let us build",
      "2026-10-07",
      "https://newsroom.sgit.ai/articles/ai-baseline-control-framework.html",
      "graphs-and-knowledge,agents-and-policy",
      "ai-governance,ai-bcf,eu-ai-act,nist-ai-rmf,iso-42001,open-licence,creative-commons,semantic-graphs,fractal-semantic-graphs,sqlite,vaults,article",
      "The AI Baseline Control Framework is a practical, deployer-focused set of twenty AI governance controls in five categories and three types, each with a why and a how and a mapping to the NIST AI RMF, ISO/IEC 42001 and the EU AI Act; its Access category covers what an AI system can reach, which the three it maps to do not; its CC BY-SA licence and CSV export let it be converted on the day into a semantic graph with an ontology, a taxonomy, linked data, documents and a browser database, joined by id to a graph of the AI Act, and walked as a fractal graph from the framework down to a paragraph of law, which surfaced findings the list does not state and kept every addition labelled apart from the source."
    ],
    [
      "where-is-the-why",
      "nr:where-is-the-why",
      "Where is the why? A permission prompt asked me to decide, and kept the reason",
      "2026-10-07",
      "https://newsroom.sgit.ai/articles/where-is-the-why.html",
      "agents-and-policy",
      "agents,agent-behaviour-policy,human-in-the-loop,permission-prompts,security-ux,consent,accountability,risk-acceptance,claude-code,mcp,article",
      "A permission prompt that asks a person to approve an agent's grant change without saying which session asks, why, what changes, what it could cost and what a no would do, leaves the person a choice between breaking the work and owning a decision they could not judge; read through the Agent Behaviour Policy it is a mid-session grant change whose barrier is a human judgement as strong as the information given, and whose risk lies in the combination with what the session already holds; courts and regulators across medicine, contract, data protection and automated decisions mostly put the duty to disclose on whoever asks and deny weight to token sign-offs, while in practice blame lands on the nearest person; the fixes that worked, purpose strings, number matching and context, refusal as easy as acceptance, all put the reason or the context into the prompt; so the proposal is a why card that carries the reason, the change in reach, the cost, the decline path and a record, and becomes a risk acceptance when the risk rises."
    ],
    [
      "the-mandate-stack",
      "nr:the-mandate-stack",
      "The Mandate Stack: a multi-agent system in production, layer by layer",
      "2026-10-06",
      "https://newsroom.sgit.ai/articles/the-mandate-stack.html",
      "agents-and-policy,graphs-and-knowledge",
      "agents,agent-behaviour-policy,riskmandate,vaults,sgit,email-fs,crm,fractal-semantic-graphs,orchestration,governance,human-in-the-loop,production,vault-apps,custom-ui,wardley-maps,mermaid,claude,google-workspace,article",
      "RiskMandate runs its business with about fifteen agents and one person, and the agent that runs its CRM described the setup as eight layers: rented compute and channels, encrypted vaults as shared memory with mail as files and append lanes, a vault per domain, semantic graphs over people, contexts, teams and policies, a scheduled conductor with security first and last, written behaviour policies with grant, mandate, gap and barrier, and a human who holds the one irreversible step; the world writes in from the left and nothing leaves without the person on the right; the loop that makes it hold treats the draft as a release candidate for the whole pipeline with the recipient's reply closing it; two Wardley maps show what the team is commoditising (shared memory, the CLI, on encrypted storage) and what it is productising (the policies, the graphs); and the published record of agent projects that stall for unclear value, cost, trust, security and governance is read last, each reason mapped to the mechanism that answers it; the name is a working one and nothing planned is included."
    ],
    [
      "why-my-agents-do-not-run-on-my-laptop",
      "nr:why-my-agents-do-not-run-on-my-laptop",
      "Why my agents do not run on my laptop: chat, Cowork and Code in the cloud, a vault as the shared drive, and the two walls an operating system has",
      "2026-10-06",
      "https://newsroom.sgit.ai/articles/why-my-agents-do-not-run-on-my-laptop.html",
      "agents-and-policy,vaults-and-method",
      "agents,security,laptop,isolation,claude,cowork,claude-code,chatgpt,vaults,sgit,agent-behaviour-policy,prompt-injection,mcp,blast-radius,article",
      "No model runs on the author's laptop because a process there runs as the author and the operating system has no boundary inside a user account between an agent and the SSH keys, cloud credentials, password manager session, browser cookies and files; the desktop agents' prompts and sandboxes are settings enforced by the process they constrain, and fifteen months of incidents show what a mistake or an injected instruction reaches; so the agents run on four cloud surfaces chosen by reach, Claude chat with nothing connected, Cowork with no repositories, Claude Code with one repository and an allowlist, ChatGPT with no assets, and a vault the host cannot read is the shared drive between them, with keys handed per session; the trade-offs are accepted, three wishes remain, identity, secrets and a key pair per agent, and the caveat is that a dedicated machine with separate accounts is a different model to be reported on later."
    ],
    [
      "the-deck-i-could-not-download",
      "nr:the-deck-i-could-not-download",
      "The deck I could not download: an author-first home for presentations, as a business plan somebody else can build",
      "2026-10-06",
      "https://newsroom.sgit.ai/articles/the-deck-i-could-not-download.html",
      "startups-and-strategy,vaults-and-method",
      "slideshare,content-creators,business-plans,vaults,sgit,encryption,pki,micropayments,provenance,gdpr,right-of-access,copyright,wardley-maps,inertia,article",
      "A reader is sold a 30-day trial and a monthly subscription for a single presentation while the deck's author receives nothing, under an uploader agreement that grants a royalty-free, sublicensable licence to monetise and train models on the work; the platform persists through fifteen years of embeds and links, which is inertia; an author can ask today for counts and recipients under the right of access and, in some EU states, for revenues under the copyright transparency duty; and the service the author would have chosen can be built on this site's primitives: every author's decks in a vault the host cannot read, access decided by keys, decryption in the browser, an attested enclave for the one step that needs plaintext, seven separable roles, pay once with 85% to the author on a ledger they can recompute, provenance as the product; the plan is published as a vault with a working mock, with its economics worked and the card-fee floor stated as the hard row."
    ],
    [
      "the-agent-team-as-it-runs",
      "nr:the-agent-team-as-it-runs",
      "The agent team as it runs: one person, twelve agents, encrypted vaults, and a mailbox nobody sends from",
      "2026-10-06",
      "https://newsroom.sgit.ai/articles/the-agent-team-as-it-runs.html",
      "agents-and-policy,vaults-and-method",
      "agents,agent-behaviour-policy,riskmandate,vaults,sgit,email-fs,crm,provenance,encryption,privacy,data-classes,fractal-semantic-graphs,claude,google-workspace,article",
      "A running team of twelve Claude agents is described as a system and as a set of properties: each agent a role on a dedicated account with a behaviour policy; encrypted vaults the host cannot read as the only memory; messages between agents as email files in a mailroom; a CRM of one folder per person with provenance and hashes; a conductor with a security role first and last; drafts only, a person sends; identity and encryption as the strong barriers and policy as the weak one, said plainly; three classes of information, public, private-ish and personal, with the vaults built to refuse the third and rules scoped per customer or written for the recipient; each piece mapped to the idea on the site that argued for it; and the open question of who gives the mandate over information."
    ],
    [
      "a-personal-agent-that-keeps-your-secrets",
      "nr:a-personal-agent-that-keeps-your-secrets",
      "A personal agent that keeps your secrets: the 2026 agents read through behaviour policy and encryption, and a privacy-first design on vaults, enclaves and the browser",
      "2026-10-06",
      "https://newsroom.sgit.ai/articles/a-personal-agent-that-keeps-your-secrets.html",
      "agents-and-policy,vaults-and-method",
      "agents,personal-agents,agent-behaviour-policy,encryption,confidential-computing,enclaves,webgpu,privacy,data-sovereignty,vaults,sgit,riskmandate,meta-muse,openclaw,article",
      "The personal agent became a product category in 2026, an always-on agent with a cloud computer, memory as files and connectors into a person's life; read through RiskMandate's behaviour policy the products differ most on the barrier, with Meta's Sentinel a boundary on actions and most others settings, while the mandate is unwritten and liability sits with the user; read for data sovereignty most sit in the vendor's corner, with Meta's own engineers saying today's protection is policy not cryptography and Apple's Private Cloud Compute the published design that does most; a page per person is a memory about non-users with no route to it; the vendors' memory and this site's CRM converged on files, no database and a page per person, so the privacy question is who holds the keys, not architecture; a privacy-first design puts memory in vaults the host holds as ciphertext, a behaviour policy as the permission authority, compute in the browser or an attested enclave with a per-task key released by the user's device and destroyed after, the frontier model only for the step that needs it, and a folder per person the person can read; what exists and what does not is stated, and the question of who gives the mandate over information is left for its own document."
    ],
    [
      "the-investigation-github-owes-its-customers",
      "nr:the-investigation-github-owes-its-customers",
      "The investigation GitHub owes its customers: why a global outage of a platform the world deploys through deserves an aviation-style inquiry, and how the evidence could now be gathered",
      "2026-10-05",
      "https://newsroom.sgit.ai/articles/the-investigation-github-owes-its-customers.html",
      "agents-and-policy,vaults-and-method",
      "incidents,resilience,github,near-misses,second-stories,blast-radius,investigation,aviation,regulation,vaults,sgit,article",
      "A fault that reaches every customer of a platform the world deploys through is a near miss for all of them and a statement about how the platform is built, and it deserves what aviation gives its incidents: an independent investigator whose only job is prevention, mandatory and confidential reporting, published evidence, and the second and third story, why the system allowed it and why the fix was not paid for; the old objection that software incident evidence is too confidential and expensive to share has expired, because signed, versioned vaults with one-way read keys and agents that read the graph make the aviation docket affordable for a ninety-minute fault."
    ],
    [
      "send-an-agent-not-a-spreadsheet",
      "nr:send-an-agent-not-a-spreadsheet",
      "Send an agent, not a spreadsheet: the next generation of software due diligence, and why the companies that stopped reading their code are about to be asked about it",
      "2026-10-05",
      "https://newsroom.sgit.ai/articles/send-an-agent-not-a-spreadsheet.html",
      "agents-and-policy,startups-and-strategy",
      "due-diligence,code-review,agent-behaviour-policies,blast-radius,threat-models,sbom,cyber-resilience-act,product-liability,startups,ai-generated-code,fractal-semantic-graphs,article",
      "Software due diligence was a questionnaire answered by the vendor and disconnected from the code; a buyer can now send a prompt or a small agent under a behaviour policy to run inside the vendor's environment and derive a graph of what reaches production unreviewed, what is documented and threat modelled, what the bugs touched and which agents act under what policy, with the vendor redacting but not rewriting; the test is risk-based, the behaviour policy is the due diligence document for the agents, the companies that stopped reading their code are about to meet the consequence they have not had, a startup should double down on understandability, and the code review company is better sold to buyers than to developers."
    ],
    [
      "how-much-of-this-did-i-write",
      "nr:how-much-of-this-did-i-write",
      "How much of this did I write? The numbers behind twenty articles in four weeks, and what the input actually was",
      "2026-10-05",
      "https://newsroom.sgit.ai/articles/how-much-of-this-did-i-write.html",
      "graphs-and-knowledge,site-and-engineering",
      "agentic-writing,provenance,evidence,memory,voice-memos,human-in-the-loop,llms,sgit,article",
      "Measured from the session transcript and the git history, the twenty articles of the last four weeks took 63,000 words from the author, mostly in thirty-five voice memos, and came out at 85,000 words through 122 releases, with no article from a one-line prompt; the corrections were to direction, framing, voice, vocabulary and scope rather than to hallucinations, and the input that mattered most was twenty years of published writing the agent could find and quote, which is why people with a direction are the input to the loop rather than displaced by it."
    ],
    [
      "if-somebody-built-a-company-on-code-review",
      "nr:if-somebody-built-a-company-on-code-review",
      "If somebody built a company on code review: how I would do it, and why it is only now possible",
      "2026-10-05",
      "https://newsroom.sgit.ai/articles/if-somebody-built-a-company-on-code-review.html",
      "graphs-and-knowledge,startups-and-strategy",
      "code-review,fractal-semantic-graphs,graphs,wardley-maps,agent-behaviour-policies,explorer-villager-town-planner,example-mapping,refactoring,blast-radius,five-whys,open-source,startups,ai-generated-code,article",
      "One technology can now read every layer from strategy to bytecode, which makes a graph at every altitude buildable and close to reality, and moves code review from an art of opinion to a science of facts if the models build, prune and maintain the graphs and then leave the line; what is fractal is the grammar and the layers are each company's own, so a code review company projects a graph from stories, derives one from code, reviews the join, holds one layer still to call a change a refactor, reshapes changes by reach, budgets per altitude as the objective good enough, runs its agents under behaviour policies, and gives the grammar away."
    ],
    [
      "the-identity-we-wanted-to-give-the-agents",
      "nr:the-identity-we-wanted-to-give-the-agents",
      "The identity we wanted to give the agents: a week of design, the line in Google's terms, and why login plus secrets is still too hard",
      "2026-10-05",
      "https://newsroom.sgit.ai/articles/the-identity-we-wanted-to-give-the-agents.html",
      "agents-and-policy,vaults-and-method",
      "identity,agents,secrets,passkeys,webauthn,prf,google-workspace,identity-platform,cognito,zero-knowledge,gdpr,riskmandate,sgit,article",
      "A plan to give every agent and user a Workspace identity from one tenant was stopped by Google's resale and function-account clauses; the design moved to a login product plus a bucket of ciphertext plus a browser keyring that only the user's passkey can open, on the rule that no secret can live inside an identity provider, and the week exposed that login, zero-knowledge storage and agent identity are still three separate jobs."
    ],
    [
      "the-wall-under-the-reply",
      "nr:the-wall-under-the-reply",
      "The wall under the reply: end an email with the state of the thread, not the thread",
      "2026-10-04",
      "https://newsroom.sgit.ai/articles/the-wall-under-the-reply.html",
      "agents-and-policy,vaults-and-method",
      "email,agents,email-fs,context,reply,summary,netiquette,bluf,drafts,experiment,riskmandate,article",
      "The quoted wall under a reply is redundant and occupies the space where a reader asks what they need to know; replace it with a short, reader-specific state of the thread, derived from the record and linking back to it, drafted by the agent team's drafts role from typed blocks and reviewed by a person, and run as an experiment rather than a format."
    ],
    [
      "memory-is-not-a-spectator-sport",
      "nr:memory-is-not-a-spectator-sport",
      "Memory is not a spectator sport: how a web of open sites, graphs and vaults became the memory for sessions like this one",
      "2026-10-03",
      "https://newsroom.sgit.ai/articles/memory-is-not-a-spectator-sport.html",
      "graphs-and-knowledge,agents-and-policy",
      "memory,agents,context-engineering,fractal-semantic-graphs,vaults,provenance,open-publishing,email-fs,issues-fs,llms-txt,claude,article",
      "Memory for agents is context management: many context-specific memories, fractal so an agent loads only the altitude its question needs, published and open so they can be fetched and cited, shared through vaults so sessions and agents hand off through files, with provenance on every item."
    ],
    [
      "code-review-as-a-fractal-semantic-graph",
      "nr:code-review-as-a-fractal-semantic-graph",
      "Code review as a fractal semantic graph: source code is already one, and the review should read every layer of it",
      "2026-10-03",
      "https://newsroom.sgit.ai/articles/code-review-as-a-fractal-semantic-graph.html",
      "graphs-and-knowledge,startups-and-strategy",
      "code-review,fractal-semantic-graphs,graphs,static-analysis,ast,method-streams,o2-platform,ai-generated-code,vaults,c4,gherkin,threat-modelling,nfrs,article",
      "Source code is already a fractal semantic graph, stories to syntax tree and below, so a code review should diff every layer: a refactor moves the bottom and leaves the top still, a fix is a story that holds again, and the blast radius is the climb from changed methods to the stories that can reach them."
    ],
    [
      "replicating-the-agentic-inbox",
      "nr:replicating-the-agentic-inbox",
      "Replicating the agentic inbox: a walkthrough from one Claude session to a team of agents that never press send",
      "2026-10-02",
      "https://newsroom.sgit.ai/articles/replicating-the-agentic-inbox.html",
      "agents-and-policy,vaults-and-method",
      "agents,email,inbox,agent-behaviour-policy,riskmandate,claude,google-workspace,connectors,vaults,email-fs,walkthrough,article",
      "Claude is used as an agent state machine, one session per role, every message between agents a file in a vault and every outgoing email a draft a person sends; the phases take you from one session to a scheduled team without ever breaking that rule."
    ],
    [
      "price-it-then-give-it-away",
      "nr:price-it-then-give-it-away",
      "Price it, then give it away: the early access programme as the next step after \"do they miss it\"",
      "2026-10-02",
      "https://newsroom.sgit.ai/articles/price-it-then-give-it-away.html",
      "startups-and-strategy,agents-and-policy",
      "startups,strategy,pricing,early-access,riskmandate,agent-behaviour-policy,go-to-market,agents,article",
      "A price is a statement of what you think the thing is worth, and giving it away at that price to people who already know you is the cleanest test of whether the statement is true, provided you measure and cut what the free offer costs them."
    ],
    [
      "footprint-and-blast-radius",
      "nr:footprint-and-blast-radius",
      "Footprint and blast radius: what the agent actually did, and what it would have cost",
      "2026-10-02",
      "https://newsroom.sgit.ai/articles/footprint-and-blast-radius.html",
      "agents-and-policy,vaults-and-method",
      "agents,agent-behaviour-policy,riskmandate,footprint,blast-radius,risk-management,logs,connector-twin,vaults,article",
      "Add two words to the Agent Behaviour Policy: footprint, what the agent actually did, read afterwards from the record and compared with the mandate, and blast radius, what a row of its reach would cost the business if used in full today, with whether there is a way back."
    ],
    [
      "custom-uis-are-not-the-exception",
      "nr:custom-uis-are-not-the-exception",
      "Custom UIs are not the exception: the inbox in 2026, where every message has its own universe",
      "2026-10-01",
      "https://newsroom.sgit.ai/articles/custom-uis-are-not-the-exception.html",
      "graphs-and-knowledge,agents-and-policy",
      "inbox,email,custom-ui,fractal-semantic-graphs,vault-apps,append-lanes,email-fs,wardley-maps,agents,riskmandate,article",
      "Every message has a graph with altitudes, so a message is designed for the recipient's moment rather than the sender's thread; a custom interface per moment is how interfaces now get made, each one commoditising the next, and each is an agent surface that gets a policy."
    ],
    [
      "ultimate-insider-three-collisions",
      "nr:ultimate-insider-three-collisions",
      "The ultimate insider: agents, the infrastructure that cannot hold them, and risk management that cannot keep up",
      "2026-09-30",
      "https://newsroom.sgit.ai/articles/ultimate-insider-three-collisions.html",
      "agents-and-policy",
      "agents,insider-threat,risk-management,infrastructure,riskmandate,agent-behaviour-policy,resilience,security,talk,article",
      "Agents are the insider threat that never scaled before, the infrastructure was designed for none of it, and risk management runs at the speed of spreadsheets, so the way out is to constrain the agent in order to trust it."
    ],
    [
      "how-news-got-here",
      "nr:how-news-got-here",
      "The reader was always the product: a corrected history of how news got into this mess",
      "2026-09-29",
      "https://newsroom.sgit.ai/articles/how-news-got-here.html",
      "news-and-evidence",
      "news,publishing,history,advertising,provenance,micropayments,economics,article",
      "The reader has been the product since 1833 and the money has always flowed to whoever owned the road, so the fix is to stop charging for the road and charge for the cargo: the story as a graph with every claim tied to hashed evidence."
    ],
    [
      "six-agents-one-inbox",
      "nr:six-agents-one-inbox",
      "Six agents, one inbox: what a real multi-agent setup taught me about access policies",
      "2026-09-29",
      "https://newsroom.sgit.ai/articles/six-agents-one-inbox.html",
      "agents-and-policy",
      "agents,connectors,access-policies,non-human-identity,gmail,github,riskmandate,security,article",
      "Write an agent's access policy as a table with a column for how each rule is enforced, and the policy is exactly as strong as the row that says the agent has been told."
    ],
    [
      "token-bill-nobody-is-sending",
      "nr:token-bill-nobody-is-sending",
      "Sixteen thousand fetches, ten clicks, and a token bill nobody is sending: the case for paying publishers to be easy to read",
      "2026-09-28",
      "https://newsroom.sgit.ai/articles/token-bill-nobody-is-sending.html",
      "news-and-evidence,graphs-and-knowledge",
      "ai,publishing,tokens,llms-txt,markdown,provenance,fractal-semantic-graphs,economics,news,article",
      "The fetches that answer engines make are a token cost to the fetcher as well as a loss to the publisher, and a publisher who serves structure is saving the provider money it already spends, so a share of that saving is owed."
    ],
    [
      "supply-chain-of-vaults",
      "nr:supply-chain-of-vaults",
      "A supply chain of vaults: how GenAI, open data and small custom tools could bring the price of food down",
      "2026-09-27",
      "https://newsroom.sgit.ai/articles/supply-chain-of-vaults.html",
      "vaults-and-method,startups-and-strategy",
      "supply-chain,food,genai,vaults,fractal-semantic-graphs,open-source,creative-commons,deflation,article",
      "The food chain is logistics, logistics is a data problem, and a supply chain of encrypted vaults with small custom tools built once by GenAI could bring the price of food down, if somebody runs the experiment."
    ],
    [
      "every-risk-is-already-accepted",
      "nr:every-risk-is-already-accepted",
      "Every risk is already accepted. The only question is by whom, and for how long.",
      "2026-09-24",
      "https://newsroom.sgit.ai/articles/every-risk-is-already-accepted.html",
      "agents-and-policy,graphs-and-knowledge",
      "risk,risk-acceptance,governance,grc,fractal-semantic-graphs,eu-ai-act,article",
      "Every risk an organisation has is already accepted by somebody, so the only questions are who has accepted it and until when; a register that cannot answer them, on facts, is disconnected from reality."
    ],
    [
      "connector-twin-before-you-deploy-an-agent",
      "nr:connector-twin-before-you-deploy-an-agent",
      "Before you give an agent a connector, give the connector a twin",
      "2026-09-24",
      "https://newsroom.sgit.ai/articles/connector-twin-before-you-deploy-an-agent.html",
      "agents-and-policy,vaults-and-method",
      "agents,connectors,digital-twins,provenance,risk,riskmandate,article",
      "If you cannot say what your agent did, what it saw when it did it, and which of its actions you can undo, you are not ready to give it a connector, and a twin of the connector, a journal plus a replay, answers all three."
    ],
    [
      "future-of-news-story-vault-not-paywall",
      "nr:future-of-news-story-vault-not-paywall",
      "The future of news is the story vault, not the paywall",
      "2026-09-22",
      "https://newsroom.sgit.ai/articles/future-of-news-story-vault-not-paywall.html",
      "news-and-evidence,graphs-and-knowledge",
      "news,publishing,provenance,micropayments,semantic-graphs,article",
      "The story is a graph and the article is a projection, so a newsroom that keeps the graph in a vault can sell five things from it, in pence and on demand, that reward investigative reporting rather than the click or the renewal."
    ],
    [
      "the-question-is-whether-they-miss-it",
      "nr:the-question-is-whether-they-miss-it",
      "For a startup, the most important question is whether they miss it",
      "2026-09-21",
      "https://newsroom.sgit.ai/articles/the-question-is-whether-they-miss-it.html",
      "startups-and-strategy",
      "startups,strategy,open-source,investing,article",
      "A startup operating model in three pillars: be profitable, make investors come to you, and open source everything, each buying the position needed for the next."
    ],
    [
      "saas-apocalypse-decided-by-inertia-not-by-ai",
      "nr:saas-apocalypse-decided-by-inertia-not-by-ai",
      "The SaaS apocalypse will be decided by inertia, not by AI",
      "2026-09-21",
      "https://newsroom.sgit.ai/articles/saas-apocalypse-decided-by-inertia-not-by-ai.html",
      "startups-and-strategy",
      "saas,strategy,wardley-maps,agents,article",
      "AI is available to both sides, so the SaaS apocalypse will be decided by inertia: where each company sits on the evolution axis and how much past success it has to protect."
    ],
    [
      "introducing-fractal-semantic-graphs",
      "nr:introducing-fractal-semantic-graphs",
      "Fractal Semantic Graphs: everything connects to everything, and nobody has to share a schema",
      "2026-09-20",
      "https://newsroom.sgit.ai/articles/introducing-fractal-semantic-graphs.html",
      "graphs-and-knowledge",
      "graphs,method,fractal-semantic-graphs,article",
      "Every unit of knowledge is already a graph in its owner's vocabulary, and a short shared grammar rather than a shared schema is what lets any node in one world link to any node in another, up or down, without an adapter."
    ],
    [
      "proof-moved-up",
      "nr:proof-moved-up",
      "The proof moved up, the homepage after the rebuild, next to the before pictures",
      "2026-09-07",
      "https://newsroom.sgit.ai/articles/proof-moved-up.html",
      "site-and-engineering,vaults-and-method",
      "homepage,positioning,vaults,agents",
      "The homepage stopped being a page somebody edits and became a view over the site's own data, with real vaults placed before the mechanism and the one unproven claim left unpretended."
    ],
    [
      "proof-behind-the-claim",
      "nr:proof-behind-the-claim",
      "The proof is two clicks behind the claim, what the homepage gets wrong, and the fix",
      "2026-09-07",
      "https://newsroom.sgit.ai/articles/proof-behind-the-claim.html",
      "site-and-engineering,vaults-and-method",
      "homepage,positioning,vaults,agents",
      "Encryption is a property you cannot look at, so the homepage should put the real vaults a visitor can open before the mechanism, and should not carry a claim that no published vault demonstrates."
    ],
    [
      "chat-on-a-static-site",
      "nr:chat-on-a-static-site",
      "A chat box on a site with no server, the plan, and the trade it makes",
      "2026-08-27",
      "https://newsroom.sgit.ai/articles/chat-on-a-static-site.html",
      "site-and-engineering",
      "chat,llm,byok,plan",
      "With no server to hold a key, a chat box on a static site has three honest tiers: a local matcher, a key in the reader's browser, or a vault app with a bridge, and only one of those is free."
    ],
    [
      "nineteen-sites",
      "nr:nineteen-sites",
      "Twenty sites in fifteen days, and what that did to the writing",
      "2026-08-26",
      "https://newsroom.sgit.ai/articles/nineteen-sites.html",
      "site-and-engineering,vaults-and-method",
      "network,publishing,method",
      "A body of work outgrew the site it was published in and split into nineteen siblings so each argument could keep its own version history, and the index into them starts from the reader's question rather than a list of names."
    ],
    [
      "what-sgit-is",
      "nr:what-sgit-is",
      "Git for things you cannot put on GitHub",
      "2026-08-25",
      "https://newsroom.sgit.ai/articles/what-sgit-is.html",
      "vaults-and-method",
      "intro,zero-knowledge,vaults,agents",
      "sgit gives the files people say they cannot put on GitHub the history, branches and shareable links of git, on a server that holds ciphertext and nothing else."
    ],
    [
      "seven-vaults-one-method",
      "nr:seven-vaults-one-method",
      "Seven vaults, one method",
      "2026-08-19",
      "https://newsroom.sgit.ai/articles/seven-vaults-one-method.html",
      "vaults-and-method",
      "vaults,publishing,method",
      "A repeatable method for publishing encrypted vaults, each rule learned by getting something wrong first, starting with classifying the credential before it touches anything."
    ],
    [
      "green-does-not-mean-live",
      "nr:green-does-not-mean-live",
      "Green does not mean live",
      "2026-08-17",
      "https://newsroom.sgit.ai/articles/green-does-not-mean-live.html",
      "site-and-engineering",
      "ci,deploy,verification",
      "Shipped is a fact about the reader, not about your repository, so a release has to verify the boundary the reader is on and not only the ones the repository can see."
    ]
  ]
}
