Spec features with AI and hand off to coding agent with ChatGPT
Author discusses a spec with an AI agent or ChatGPT, then hands it to an agent to code, and tests the outcome instead of reading every line.
Done with Claude Code
What they did
The author first talks through a feature with an AI agent or ChatGPT to produce a thorough spec. That spec then goes to a coding agent, which writes the code. The author doesn't read every line of the output; instead they test the result thoroughly themselves.
How it went
The author says they ship better-quality applications much faster, and that the AI handles edge cases and best practices they might have missed. Bugs that remain, like unexpected data, are ones they'd have written too.
Worth knowing
Since review happens through testing rather than reading code, the quality of your tests, especially with unexpected input data, decides whether bugs get caught.
Try it yourself with Claude Code
I want to add [feature] to this project. First interview me with questions until you can write a clear spec, and save it as SPEC.md. Then implement it and write tests that check the behavior. Finished means the tests pass and you give me a short list of steps to try the feature myself. Ask before deleting files or changing the database.
Source on HN - Feb 16, 2026
Read the original