Theoretical predictions on the sample efficiency of training policies and activation monitors
Published on January 10, 2026 11:50 PM GMTI'm worried about AI models intentionally doing bad…
Category: Opinion
If AI alignment is only as hard as building the steam engine, then we likely still die
Published on January 10, 2026 11:10 PM GMTCross-posted from my website. You may have seen…
Possible Principles of Superagency
Published on January 10, 2026 9:00 PM GMTPrior to the era of superintelligent actors, we’re…
The Case Against Continuous Chain-of-Thought (Neuralese)
Published on January 10, 2026 8:32 PM GMTMain thesis: Discrete token vocabularies don't lose information…
The false confidence theorem and Bayesian reasoning
Published on January 10, 2026 5:14 PM GMTA little backgroundI first heard about the False…
A Proposal for a Better ARENA: Shifting from Teaching to Research Sprints
Published on January 10, 2026 4:56 PM GMTTLDRI propose restructuring the current ARENA program, which…
Are there any extremely strong arguments that Acausal extortion is ineffective?
Published on January 10, 2026 1:37 PM GMTThe topic of acausal extortion (particularly variants of…
AI Incident Forecasting
Published on January 10, 2026 2:17 AM GMTI'm excited to share that my team and…
Finding high signal people – applying PageRank to Twitter
Published on January 10, 2026 2:21 AM GMTCross post, adapted for LessWrongSeveral challenges add friction…
Moral-Epistemic Scrupulosity: A Cross-Framework Failure Mode of Truth-Seeking
Published on January 10, 2026 2:24 AM GMTCrossposted from https://substack.com/home/post/p-183478095 Epistemic status: Personal experience with a…