01
Orchestrate
Workbench
An ultra-lightweight CLI that turns a plan into parallel, isolated work across AI coding agents. Dependency-aware waves, git worktrees, flexible agent profiles, review, and merge automation keep complex delivery legible.
Open source & experiments
Tools for
serious AI work.
Projects shaped by the operational problems behind modern AI systems: coordination, visibility, evaluation, and reliable delivery.
Plan→Run→Learn→Improve
02
Observe
thirdeye
Local-first observability for Claude Code, Codex, and Cursor. Capture structured session events on your filesystem, search and tag the work, inspect token usage, run evaluations, or mirror traces to Pydantic Logfire.
03
Evaluate
Valcore
A focused, local-first workbench for authoring agent judges, building datasets, and running agent evaluations from a visual UI, the terminal, or CI.
The lab
More things I've built.
Smaller products, reference implementations, and tools for learning in public.
Need help building the system behind your AI product?
Explore consulting ↗