Charting paths to less risky, more governable AI via provable safety properties at Atlas Computing 
(🦋 @miyazono.bsky.social)