I build AI systems that improve themselves under human oversight.

Cognitive AI that learns your judgment over time

• • Belief graph with contradiction detection
• • Self-improvement loop (metrics-driven)
• • Bounded autonomy: human merge gate
• • 31 MCP tools · 762 tests · running in prod

Governed build pipeline for any repo

• • One-command setup: autoloop init
• • Serial dispatch + dependency ordering
• • Config-driven: works on any language/toolchain
• • Protected paths: loop can't modify itself