Open-source tools and experiments for AI-assisted software development and developer experience
I build small, focused tools that fit into the workflows of engineers using AI coding agents. These are open source under the MIT license. More projects are landing here over the next few weeks.
A completed experiment in using specialized, in-context guidance to improve agentic code reviews. Includes 23 reviewer skills and retained real-code evaluations; the results found no conclusive improvement over ordinary agent reviews.
Browse and export AI coding assistant conversation history across Claude Code, Codex CLI, and Gemini CLI. Reaches local, WSL, Windows, and SSH homes; produces listings, markdown exports, and usage metrics.
Preflight a repository before AI-assisted refactors. Finds duplicate-code candidates, repeated documentation sections, and docs IA overlap so an agent can target real consolidation work.