Back then. (6 months ago? lol) another process messing with your file would be Bad News, today it’s just my buddy claude doing its thing

I do see most people like the idea of reading the “chat” that generates the code but idk… isn’t that just more LLM-generated slop to read?

I don’t, but maybe because I don’t personally care about prompting at all and don’t put effort on building nice prompts.

I do expect a good summary of the changes and how can I easily verify/test it. The summary or the testing guide being too vague are strong smells that the patch is slop.

computer science is when you write “Alice” and “Bob” and “Eve” on a whiteboard and the more times you write it, the more science the computer is

10-20%, and mostly because it’s some small adjustment/dumb refactor and calling Claude would waste my precious tokens 😭

and if you pair with like tangled.org that has support for change IDs, it’s *chef kiss*

Nice! Will try it tomorrow