Real example · Build apps (vibe coding)
Fire off code research jobs to async agents and read the results later
Done withOpenAI Codex Also used: Claude Code
What they did
Simon Willison ran two or three short research prompts a day on fire-and-forget coding agents (Codex Cloud, Claude Code for web, Jules, Copilot coding agent) in a throwaway GitHub repo, for example benchmarking Python Markdown libraries and compiling a C extension to WebAssembly.
How it went
About two weeks in he had 13 projects. The WebAssembly build only worked after he pushed the agent past an early give-up, and the tag-suggestion run skipped the requested visualization. He called the unreviewed reports "total slop" he would not publish unedited.
Worth knowing
He runs this on fresh, non-sensitive repos with open network access; locked-down networking is the default for private code, and an agent failing to find a solution does not prove none exists.
Try it yourself with OpenAI Codex
Replace anything in [brackets] with your details.
Research task, write code to answer it: [question, e.g. which of these three libraries is fastest at parsing Markdown?]. Install what you need, run real benchmarks, and put the code, raw results and a short report with a chart in a new folder named [folder]. Say clearly what you could not verify.Read the original
Code research projects with async coding agents like Claude Code and Codex ↗
simonwillison.net · · Personal post
Attributed to the source linked below; not our own test.
Share Share on X Share on WhatsApp
