Thinking Machines Lab sets out its worldview: 'The future worth building is human'

On 10 July 2026, Mira MuratiDealroom has a profile for this one. Try Dealroom → shared the worldview behind the mission of Thinking Machines Lab in an essay titled "The Future Worth Building Is Human".

The core thesis: the mission is to build AI that extends human will and judgment, rather than replacing it. Human values and knowledge don't average out and can't be centralised — the good future has many AIs, raised in different places and shaped by the people they serve.

Against centralisation: most AI today is trained in a handful of places and then frozen, unshaped by the people it serves. Drawing on Hayek and Polanyi, the essay argues that productive knowledge is tacit, local and dispersed — so for AI to benefit from distributed knowledge, it must itself be distributed.

Technical directions: training frontier models with multimodal interaction and customizability; tools (Tinker) that let people fine-tune model weights on their own knowledge; interaction models that widen the human–machine communication channel; and open research.

Decentralised alignment: a single locus of value alignment becomes a locus of power to be captured. The lab envisions alignment as a feature of an ecosystem of AIs — raised in different places, disagreeing, competing and learning from each other — rather than one centrally-aligned model. "We believe in keeping the weirdness alive."

Read more: Thinking Machines Lab · Mira Murati on X