lain on Nostr: Here's something I tried and it actually works: 1. Ask a reasoning LLM (i used grok ...
Here's something I tried and it actually works:
1. Ask a reasoning LLM (i used grok 3) to plan an MVP for whatever you want. I chose a server that re-encodes podcast feeds to opus, ad that has a web frontend to add podcasts. Tell it to make each step and easy '1 point story', and tell it to make it testable.
2. Save that as plan.md in a directory.
3. Start claude code in that directory, tell it go through each step and ask you after each step for a code review.
4. Do that until it's finished.
It really did produce a tested working MVP according to the specs in ~1 hour, without me having to write a single line of code.
1. Ask a reasoning LLM (i used grok 3) to plan an MVP for whatever you want. I chose a server that re-encodes podcast feeds to opus, ad that has a web frontend to add podcasts. Tell it to make each step and easy '1 point story', and tell it to make it testable.
2. Save that as plan.md in a directory.
3. Start claude code in that directory, tell it go through each step and ask you after each step for a code review.
4. Do that until it's finished.
It really did produce a tested working MVP according to the specs in ~1 hour, without me having to write a single line of code.