"Don't think about pink elephants." Humans can't seem to avoid certain thoughts. What about LLMs? Can we robustly monitor LLM activations to catch bad thoughts before they become actions? To study this, we crafted a real jailbreak causing this LLM activation scan. Details 👇

A simple AGI safety technique: AI’s thoughts are in plain English, just read them We know it works, with OK (not perfect) transparency! The risk is fragility: RL training, new architectures, etc threaten transparency Experts from many orgs agree we should try to preserve it:

2015: RNNs hallucinate LaTeX that almost compiles. 2025: Gemini 2.5 Deep Think is an IMO medalist. The models leveled up, and our safety testing did too.