I am building a research agenda tackling catastrophic risks from AI, which I am documenting on my substack Working Through AI. Where it feels suited, I'm crossposting my work to LessWrong.