We believe Control of AI should be in the hands of people building the applications. That’s why we're building the rails for switchable intelligence
Introducing Thesean with $30M to build best execution for LLMs.
We're ex-Anthropic, ex-DeepMind, and ex-quant. Our goal: maximize intelligence per dollar. Not just a model or a router, but every conceivable way to optimize AI usage.
We just made Opus 50% cheaper: thesean.ai/odyssey
The most important questions in AI are economic. How will we fund the data center buildout? How will companies make AI profitable? How will the gains from AI be distributed? Best execution is the answer to these questions. It's the optimization pressure for the AI ecosystem.
In our view this is not "distill models" or "throw some models into a fusion API". Solving it requires doing serious research.Internally, we've built models that dominate the existing Pareto Frontier. We're working with application layer companies doing quadrillions of tokens to prove out our claims.
We'll be releasing our system behind public endpoints and publishing research on how we're doing it.
Latest Research