Thinking Machines Lab Wants AI Shaped by the People Doing the Work
- Agents
- Capital, Markets, And Business Models
- Policy, Governance, And Geopolitics
The mission of Thinking Machines is to build AI that extends human will and judgment.
Recap
When a chef crafts a new recipe, it's rarely a static process: tasting, adjusting, and learning what works in that particular kitchen, with the locally available foods, people, and culture. They are pursuing a complex set of goals—a symbiosis of different nodes of knowledge, continually updated through interaction with their work and the world over time.
Thinking Machines are doing some really interesting work. They diverge from the other major labs pursuing autonomous agents, working toward a more continuous collaborative partner.
Ideas
- AI Needs the Knowledge Made Inside the WorkBringing intelligence to knowledge — distributed organizational knowledge
Thinking Machines argues that much organizational knowledge is tacit and continuously made in work, so AI should help organizations adapt models around that knowledge rather than replace them with one standard system.
- Human Participation Is an Interface and Evaluation ProblemHuman participation is a technical challenge — interaction models
The authors argue that a chat box and delayed responses cannot carry continuous human judgment, so AI systems should be built for live multimodal collaboration and evaluated for joint human–AI outcomes rather than autonomy alone.
- Autonomy Benchmarks Do Not Measure the Whole JobHuman participation is a technical challenge — evaluation target
The article says autonomous task-completion measures can track an important capability, but do not measure whether people and AI together make better decisions or build knowledge inside real organizations.
- Centralized Alignment Concentrates PowerDecentralized alignment — values and power
The authors contend that when a small number of labs determine a model’s values and voice, alignment becomes a concentrated power center rather than a process shaped by the people affected by the system.
- Ownership and Customization Change the Lab’s IncentivesHuman participation is a technical challenge — long-run incentives
Thinking Machines argues that a lab selling one standard model benefits from absorbing what makes customers distinct, while a lab focused on customization benefits when customers retain and build on their own advantages.
Tags
- Benchmarks And Evaluation
- Human-In-The-Loop Agents
- Enterprise AI Adoption
- Post-Training
- AI Policy And Governance
- Frontier Lab Business Models
- AI Economics