Lesser Wrong Feed

 "We've been fine before, so we'll be fine again" is a fallacy (in the more dangerous direction).
 Thu, 19 Mar 2026 04:57:11 GMT
 The "talker–feeler gap": AI valence may be unknowable
 Thu, 19 Mar 2026 04:30:04 GMT
 Latent Reasoning Sprint #2: Token-Based Signals and Linear Probes
 Thu, 19 Mar 2026 03:39:32 GMT
 An interactive version of the extropians mailing list
 Wed, 18 Mar 2026 22:17:23 GMT
 Metagaming matters for training, evaluation, and oversight
 Wed, 18 Mar 2026 21:26:35 GMT
 Formal Verification: The Ultimate Fitness Function
 Wed, 18 Mar 2026 22:59:53 GMT
 Training on Documents About Monitoring Leads To CoT Obfuscation
 Wed, 18 Mar 2026 20:51:08 GMT
 LessWrong's UX may not be living up to its ideas
 Wed, 18 Mar 2026 20:16:22 GMT
 Two Skillsets You Need to Launch an Impactful AI Safety Project
 Wed, 18 Mar 2026 20:09:37 GMT
 Anthropic vs. DoW #5: Motions Filed
 Wed, 18 Mar 2026 19:00:20 GMT