Lesser Wrong Feed

 Deployment Awareness Matters More Than Evaluation Awareness
 Fri, 26 Jun 2026 22:54:06 GMT
 Just a Wrapper? How Much Do Scaffolds Matter?
 Fri, 26 Jun 2026 22:21:40 GMT
 What did "scheming" and "mech interp" mean pre-2023?
 Fri, 26 Jun 2026 22:09:35 GMT
 Why are adversaries assumed to be incapable of responding to AI risk?
 Fri, 26 Jun 2026 21:51:14 GMT
 Screencasts could be scalable data + evals for single-user emulation (Guardian Angels)
 Fri, 26 Jun 2026 21:12:19 GMT
 Randomness Testing with Bayesian Stats
 Fri, 26 Jun 2026 20:28:30 GMT
 Should we combine protocols for AI Control Research?
 Fri, 26 Jun 2026 19:02:53 GMT
 Not making a strong argument is a relief
 Fri, 26 Jun 2026 18:41:39 GMT
 The Case for Model Forensics
 Fri, 26 Jun 2026 15:09:10 GMT
 White House Will Ad Hoc Decide Who Can Individually Access GPT-5.6
 Fri, 26 Jun 2026 14:53:09 GMT