I’m an independent developer building tools that make AI coding agents more durable, observable, and accountable.
My work sits at the intersection of developer tooling, agent orchestration, and evidence-driven engineering—turning fragile one-off prompts into repeatable systems with persistent state, regression checks, and operational guardrails.
|
A self-improving operating layer for AI coding agents, with versioned skills, hooks, regression evals, and a calibration ledger. Python · PowerShell · TypeScript |
Mission control for parallel CLI coding agents running in isolated Git worktrees, with terminals, diffs, and live monitoring. TypeScript · React · Playwright |
A codebase intelligence tool that maps symbols, calls, imports, hotspots, and overlap into an interactive graph. JavaScript · Go · GitHub Actions |
- Goal Prompts — 30 evidence-driven mission briefs for auditing repositories and producing actionable reports. Try it live →
- Engineering Board — autonomous work routing, validation, and session-end accountability for Claude Code. View the board →
- ViewForge — a tested, learning-loop pipeline for research, scripting, motion, editing, distribution, and measurement.
- Persistent state over ephemeral context — important work survives restarts, handoffs, and context limits.
- Measurable learning over vague “self-improvement” — predictions, outcomes, and regressions stay on the record.
- Isolation and recovery by design — worktrees, bounded loops, deterministic gates, and inspectable artifacts.
- Tools that compose — focused building blocks for real agent-assisted engineering workflows.
TypeScript · React · Next.js · Python · PowerShell · Shell · GitHub Actions · Playwright
Building practical infrastructure for serious agentic engineering.