Suggest a source
Proof-of-concept stage. The Compendium repo is currently private and pull requests are not open. The form below sends your suggestion to the maintainer’s private review queue; you won’t see a public PR thread, but the suggestion is logged and reviewed on the next weekly sweep.
Found an AI safety paper, post, or blog the compendium should cover? Drop the URL below. Each submission lands in the maintainer’s private intake (raw/inbox/_suggested/) for review against the compendium’s scope.
What happens after you submit
- The suggestion is filed into the private intake folder on a new branch.
- The maintainer reviews suggestions on the next weekly sweep (Monday 06:00 UTC).
- If the source is in scope and not already covered, it gets added to the weekly review queue and eventually compiled into a summary page — typically within 1–2 weeks.
- Because the repository is private during the proof-of-concept stage, you will not see a public PR or issue thread. If you provided an email above and want a confirmation, mention that in the note.
The compendium covers AI safety, broadly construed — technical research (alignment, interpretability, evaluations, agentic safety, scalable oversight), governance and policy work, risk and capability assessments, and field-shaping commentary. Out of scope: pure capabilities work without a safety angle, news commentary without primary sources, and speculation without empirical or formal grounding.
Other ways to send feedback
For corrections, scope questions, or anything that doesn’t fit the form above, email kevin@itforhumanity.be. Email is the only other feedback channel during the PoC stage.