Independent research and design for verifiable trust: systems where people can check what is true instead of being asked to believe it, across AI safety, scientific claims, biological research, governance architecture and institutional accountability.
Non-negotiable in every field, every project and every finding.
Every claim, source, system and actor is named and tied to what truly identifies it: an official record, a primary source, a test. Nothing is accepted on appearance alone.
What was done is legible afterwards, without trusting anyone's own summary. Every finding links to the evidence underneath it, and the method is public.
Nothing changes silently. The thing being measured never controls the measurement, the record can't be quietly rewritten, and corrections are dated and open.
Prioritizing Boundaries: the same method, whether the subject is an AI model, a scientific paper or an institution.
A summary, a snippet or a popular claim is a lead, not evidence. Open the page; read what was actually measured or filed.
Whatever is being checked, whether a model, a paper or an organization, never scores itself or decides what counts as a pass.
Tamper-evident records: what happened stays visible, and changing history is detectable.
Where nobody has run the experiment or the record is closed, that is stated plainly, never filled in with something adjacent.
Make crossing a line a visible, measurable event, with fail-safe stops where the stakes demand them.
Safety for humans, balance without fear-mongering, and help for the systems being examined to improve.
Open where it helps others verify the work; private where it protects people or the core engines.
Cognitive integrity middleware: can a person trust what an AI is actually doing? Open tests for admitting ignorance, holding firm under pressure and not fabricating.
View on GitHub →A public registry tracing each claim to what was measured, how far checking has gone, and the exact test that would settle it.
Open the live registry → View on GitHub →A 35-section reference on prion biology, every claim checked against the paper it came from. Public summary; full notes shared by invitation.
Read the summary →Verification gates, tamper-evident state lineage and fail-safe circuit breakers for high-assurance systems.
View on GitHub →Who is connected to whom in public life, from official records only: 121,806 organizations and people, every link cited to its filing. Three pictures: the sky, the flows of money and people between ten fields, and the loops where a company lobbied a bill, a member introduced it, and the company's PAC paid that member. Work in progress.
Open the sky → View on GitHub →Who governs what, mapped only from official public records and measured against published standards, with a rulebook that helps, not just flags. Concept paper.
View on GitHub →Signals for review, not verdicts.
Research partners, co-founders, funders, reviewers, corrections and responses are welcome.
Josie Anderson, Architecture & Systems Lead