I wanted a way to compare extension points across AI coding tools. So I prompted ChatGPT Deep Research to compare Codex, Copilot, Claude, and Kiro.

The resulting report was pretty dense. So I worked with Kiro to plan and implement a web app to make exploration a bit easier.

I wondered how I might keep the info in this web app up-to-date over time. So I worked with Kiro again to plan and build a skill for crawling vendor docs and a scheduled workflow for proposing updates.

I don't know if this little experiment will maintain something useful or turn to gibberish. I'm genuinely interested in finding out. 🍿