Software engineer building local-first agent harnesses, multi-agent orchestration loops, and LLM diagnostic tooling.
Creator of VibeHarness EasyArchitect
Orchestration fable-sonnet-orchestrator-kit claude-upgrades
Diagnostics codex-reasoning-bug-516-token codex-reasoning-fix-skill
Foundations Projects (Python)
I enjoy designing small, inspectable tooling around AI coding agents—from grammar-constrained execution loops for 3B local models in Ollama to deterministic Stop hooks that enforce real pull-request review gates. You can browse my open-source projects here.
I publish technical notes on my blog covering two-phase constrained decoding, multi-agent worktree orchestration, and empirical transcript analysis of LLM reasoning collapse. Everything is built to stay lean, reproducible, and verifiable from outside the model.
Find me on