Full-stack builder — developer tools & infrastructure for AI coding agents.
I build tools that make coding agents genuinely useful: code intelligence they can query instead of guess with, and video understanding they can cite by timestamp. Most of my product work ships in private repos; the tools below are the public side.
| Project | What it is | Stack |
|---|---|---|
| codelens-mcp-plugin | Rust MCP code-intelligence daemon for coding agents — live indexes, bounded hybrid retrieval, call-graph & impact analysis, host-neutral workflows | Rust · tree-sitter · LSP · MCP |
| timecode-agent | Transcript-first video understanding for coding agents — timestamped evidence ledgers, lazy visual verification, editorial handoff (EDL / OTIO / FCPXML / SRT) | Python |
| homebrew-tap | Homebrew formulae for my CLI tools | Ruby |
- 🤖 Agent infrastructure — MCP servers, code-intelligence backends, and multi-agent harnesses that keep long-running agents grounded in evidence
- 🔍 Code intelligence — live symbol indexes, hybrid retrieval, call-graph and blast-radius analysis tuned for agent token budgets
- 🎬 Video understanding — transcript-first indexing with timestamped evidence and clean handoff into editorial formats
More about how I build
- Agent-first development: coding agents as implementer + verifier, with deterministic gates (build, test, lint) before anything counts as done
- Evidence over vibes — the tools I build exist to keep agents answering from indexes and ledgers instead of guesses