Self Play Labs

We’re a tiny research lab working on generalization, reliable acceleration through recursively self-improving systems, and sample-efficiency. We ship the RSI mechanisms and systems we build as live deployments, and make them available as frontier research substrate.

We’re building toward the inflection point where AI agents actively write, benchmark, and verify the code of their own foundation architectures. Past that point, each generation of the system can propose, test, and ship improvements to the next, opening an autonomous self-upgrade cycle that compounds over time.

If any of this is interesting, reach out at research@selfplay.computer