One purpose per policy
A policy gives one or more agents a job. It defines when they start, which tools they can use, and what done looks like.
Build systems
that evolve.
ReSM is networking tooling for agent systems. Give agents a role, connect their policies into chains, and improve each part over time.
From proteins
to systems.
A protein has a job. It responds to what’s around it and changes state as it works. Put enough of these little state machines together and you get tissue, then an organism. We think agent systems can grow the same way.
Connect policies into chains
A result from one policy can start the next. Progress lives in shared state, so the work can pick up again when a session ends.
Build chains out of chains
Combine chains around a larger goal. When a better model, harness, or policy comes along, swap it in and keep going.
We’re starting with Wittle Rooms.
We want to build a product people enjoy using that can pay for itself. Shipping features, learning from user feedback, and watching cloud spend are different jobs. Each can have its own policy, and we can steer them together toward that goal.