I build AI agent infrastructure.
My main work is RCOS — Recursive Capability Operating System: a system for routing goals into durable execution, verifying results with evidence, and reusing capabilities that pass explicit evaluation.
goal → workflow manager → execution → evidence → evaluation → reusable capability
Current focus:
- durable agent execution and recovery
- capability acquisition, evaluation, promotion, and reuse
- evidence-backed receipts instead of self-reported success
- DSH / Archon integration
- operator tooling for inspecting and controlling agent work
The goal is simple: make agent systems more capable without making their behavior less inspectable.