Transparency

Every entry is researched against primary sources and needs at least 3 independent credible ones before it is published. 2 of 226 currently fall short and are held back — listed openly on the audit dashboard rather than hidden. No self-reported data.

An AI agent scans primary sources — company websites, EU project databases, arXiv, and major tech press — every week to flag stale data and propose updates for human review.

There is no confidence grade to interpret. Each entry shows what it rests on: how many independent credible sources it carries, and when each was last accessed. An entry that drops below the bar leaves the listings until the sources are found — the bar is never lowered.

Full transparency means every change is attributed — AI suggestions are logged, human decisions are recorded in git history. The complete audit trail is public and permanent.

Each weekly sweep takes a fixed slice of the directory and re-reads every source on it, so the whole directory cycles roughly monthly. Each run is dated and public, including the entries it could not complete.

The curation protocol, source requirements, confidence rubric, and audit schedule are all documented and public. Anyone can read exactly how this directory is built and maintained.