We received a $20,000 Claude usage credits award through the MIT–Anthropic AI for Science study.

It lands where most of our compute for agents goes. Much of our agent stack runs on Claude as the base model, including the specialist agents inside the Autonomous Research Compiler, which spend a lot of tokens reading literature, drafting hypotheses, and rewriting proposals that the physics verifier sends back.

Thanks to Anthropic and the MIT team for supporting the work.