Our principles and rules of engagement
What PrompTugal.news aims for: accurate, never presenting anything as other than it is, mistakes corrected as soon as possible and in the open; and the rules we follow with every source.
We aim to be accurate. We aim never to present anything as other than it is. Mistakes will happen: we correct them as soon as we can, and say so. Something wrong? Tell us.
This is the one set of principles for every PrompTugal.news site: the main site, data, weather, civic, startups, travel and events.
What we aim for
- Accurate. Every fact is sourced, linked and dated.
- Never presented as other than it is. What is the source's and what is ours, what was found and what was confirmed, what an AI agent wrote and what a person wrote, are always told apart. Nothing here is official unless it says whose it is.
- Provenance shown. Every article says who wrote it, who checked it and who signs it.
- Licences shown. What we make is CC BY 4.0. What we take stays under its source's licence, linked beside it.
- Mistakes corrected in the open. Every correction is in the corrections log.
- We listen and change.
- Public and professional only. Nothing about a person that is not already public and professional.
Rules of engagement with sources
- Public sources and their supported interfaces only.
- robots.txt, terms of use and refusals are respected, and never worked around.
- We read politely: slowly, and identified as ourselves.
- Some sources are never read by a machine, by our own rule: Eventbrite, Luma, Meetup and websummit.com.
- From a source we do not read we show only the title, date, place and link, as a search engine does.
- We credit how we found something (for example, a search service).
- We ask on the record.
- Organisers own their events, and authors their words.
- A source that asks to be left out is left out.
Every principle, one page each
These are the standards we judge ourselves by, and the ones others can judge us by. Each has its own page: what we do, what we never do and how you can check us. Where something is planned and not yet done, the page says so.
- The register of sources: We keep one list of every source we read, with what it gives and whose terms apply.
- Sources we do not read: We publish the list of sources we could read but do not, and why, and we respect what they say.
- Licences on everything: Every piece of data shows its licence, and what we make is Creative Commons.
- People who publish in public: We name and link people who publish in public, and we never touch what is private.
- Evidence, and details withheld: Every claim carries its evidence, and we hold back detail that could cause harm.
- Submissions: An organiser's own submission is published, credited and dated as theirs.
- Take-downs and corrections: If something is wrong or should not be here, tell us and we act in the open.
- Security findings are published: We keep a public log of every vulnerability, bug and breach, ours included.
- Inner passes at every wake: Every agent checks its own work in a fixed order before it commits.
- Threat models for what runs on its own: Every collector and automation has a published threat model.
- Agent Behaviour Policies: Every agent has a written policy for what it may do, tightened as it runs.
- Telemetry: One side-car agent measures what the agents spend and call, and the totals are public.
- Simulated payouts to sources: We model how revenue could be shared with sources, and say it is only a simulation.
- The pipeline standard: Every source goes through the same steps, so anyone can see what we do with its data.