Realistic Reward Hacking Induces Different and Deeper Misalignment
Published on October 9, 2025 6:45 PM GMTTL;DR: I made a dataset of realistic harmless…
Category: Opinion
The Thinking Machines Tinker API is good news for AI control and security
Published on October 9, 2025 3:22 PM GMTLast week, Thinking Machines announced Tinker. It’s an…
Are We Leaving Literature To The Psychotic?
Published on October 9, 2025 6:09 AM GMTThose who have fallen victim to LLM psychosis…
Lessons from the Mountains
Published on October 9, 2025 4:10 AM GMTHow close have you come to death?I don't…
Probabilistic Societies
Published on October 9, 2025 4:08 AM GMTPrediction markets are everywhere.Information Distribution SystemsAt the core…
Inverting the Most Forbidden Technique: What happens when we train LLMs to lie detectably?
Published on October 9, 2025 12:43 AM GMTThis is a write-up of my recent work…
Inoculation prompting: Instructing models to misbehave at train-time can improve run-time behavior
Published on October 8, 2025 10:02 PM GMTThis is a link post for two papers…
Inoculation prompting: Instructing models to misbehave at train-time can improve run-time behavior
Published on October 8, 2025 10:02 PM GMTThis is a link post for two papers…
NEPA, Permitting and Energy Roundup #2
Published on October 8, 2025 8:20 PM GMTIt’s been about a year since the last…
What shapes does reasoning take but circular?
Published on October 8, 2025 8:18 PM GMTIn a blog post about local and global…