AI safety

Research, projects and notes on alignment and oversight. Each piece says how sure I am.

1. Introducing CruxAn early concept for checking alignment research done by AI, before anyone relies on it.project · Oct 18, 2026
2. The Atlas research agendaRobust Superalignment first, starting with Latent Invariant Alignment. Then automated alignment research, then scalable oversight.research · Jul 2, 2026