A Framework for Eval Awareness
Published on January 23, 2026 10:16 AM GMTIn this post, we offer a conceptual framework…
Category: Opinion
All Of The Good Things, None Of The Bad Things
Published on January 23, 2026 9:50 AM GMTThere’s a trap that I think many smart…
Are Short AI Timelines Really Higher-Leverage?
Published on January 23, 2026 7:28 AM GMTThis is a rough research note – we’re…
Principles for Meta-Science and AI Safety Replications
Published on January 23, 2026 6:59 AM GMTIf we get AI safety research wrong, we…
Does Pentagon Pizza Theory Work?
Published on January 22, 2026 7:24 PM GMTAs soon as modern data analysis became a…
The phases of an AI takeover
Published on January 22, 2026 7:09 PM GMTThis is a cross-post from my Substack, Clear-Eyed…
Will we get automated alignment research before an AI Takeoff?
Published on January 22, 2026 5:46 PM GMTTLDR: Will AI-automation first speed up capabilities or…
How Could I Have Learned That Faster?
Published on January 22, 2026 5:35 PM GMTThis is a question, but also a linkpost…
AI can suddenly become dangerous despite gradual progress
Published on January 22, 2026 4:47 PM GMTIn the Sable story (IABIED), AI obtains dangerous…
Releasing TakeOverBench.com: a benchmark, for AI takeover
Published on January 22, 2026 4:34 PM GMTToday, PauseAI and the Existential Risk Observatory release TakeOverBench.com:…