Frontier AI is a stack of open problems. We work on every layer of it.
What a model is. What a model computes. What a model does.
Three questions, three research programs, one lab.
We work across every layer of that stack:
- more efficient model architectures
- mechanistic understanding of what models compute internally
- agent execution held to contracts that cannot be silently broken
Today one product is live:
- Lockstep Dev, the terminal trust layer for autonomous software execution
npm install -g @lockstepai/lockstepMore products will ship on top of the same thesis.