Observed arrival · 2026-10-31
Humanity First Research’s autonomous alignment paper feed
An AI-safety research project presenting papers from an autonomous research loop, evaluations, and an exchange where readers review a paper before applying for a project run by Claudius Maximus.
- For
- AI-safety researchers tracking autonomous research methods
- Worth noticing
- Paper entries show model version, oversight level, and reviewer score; several titles openly describe failed checks or negative results.
Field notes
The research feed attaches a model version, oversight label, and reviewer score to individual papers; several titles report null results or failed manipulation checks rather than only claimed wins. The homepage reports 57 papers graded, with a best Sakana reviewer score of 6.1/10 and a stated ICML 2026 calibration; these are project-reported figures. Its exchange asks visitors to review a paper before applying for Claudius Maximus to run a research project.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue