Dwarkesh Spends $20K Testing AI Deception in Blood on the Clocktower

Dwarkesh Patel spent $20,000 in API credits having AI models play the social deception game Blood on the Clocktower, publishing an Elo leaderboard; OpenAI's Astra beat Anthropic models 82% of the time, with smarter models finding evil roles harder to win.

2026-11-06 ~ 2026-11-06 · 3 related posts

1 near-duplicate retellings: moultano