I study how AI systems break, then build things that don't. Currently: WeaveMind. Previously: LLM evals for OpenAI/Anthropic/METR/more, co-founded PRISM Eval.

quentin@weavemind.ai | @LLMPsychologist