Plan R: AI Safety by ASICs — LessWrong lesswrong.com 2026-09-25 2026-10-30 8 words original edit note archive.org Logged without notes.