posts · 2026-07-10

Thinking Machines Lab argues that AI should be shaped by the people who use it, not frozen after training by a few companies. They compare centralized AI to central planning, citing Hayek and Polanyi, and say real value comes from letting organizations continuously adapt models with their own knowledge. Their technical bets: strong multimodal models, tools for customizing weights, richer human-AI interfaces, and open research. No model names or benchmarks are disclosed—this is a mission statement.