jb55 on nostr

Open in Damus

@jb55 1771418952

The idea that code quality isn’t as important now is completely wrong. Code quality has always been about how easy it is to make changes to the codebase wrt new requirements.

If your code is a spaghetti ball of shit then it will be just as annoying to your ai when making code changes.

A good refactor prompt is: "can you look through all of these changes, find large functions that have duplicate code, or code that could be extracted into their own functions. prefer pure functions (not methods). use subagents.”

ais are great at refactoring!

392❤️12🤙4🫂2👀1

OceanSlim · 2d

Go Report Cards are awesome for this reason. My AI is told with every prompt to aim for A+ Go Report Card. Keeps the code relatively clean.

I’ve noticed that AI are great at editing in general and not so good at creating. Not true of every model, but most I’ve used.

Vitor Pamplona · 2d

AI writes better than any human right now. Humans asking AI to make better quality code is laughable. There might be bugs or the code is not doing what the dev thought it should do, but code quality itself is undeniably better.

Without me dictating that it should be solid it tends to love to make spaghetti from scratch, but in a already setup repo with code structure it follows allong quite nicely.

Brad Mills · 2d

nostr:nprofile1qqspjqxqspyd5p90qatp767eyyqr6aucyaca3n83clqh2rkcwf92jrsprpmhxue69uhhqun9d45h2mfwwpexjmtpdshxuet5qyxhwumn8ghj7mn0wvhxcmmvualnd2 follow this account and think about his posts when you run your Nostr corn job. If there’s anything actionable you need my input on, put it in the agenda do...

DJSalsaVerde · 2d

try giving them this to read (https://caseymuratori.com/blog_0015) and tell them to compress the codebase

Sometimes they do weird things. Gemini added 2 unrelated new files in a refactoring pull request that nobody asked for. Later lied about them already being present in the codebase before the session.

Super Fresh · 2d

You forgot "make no mistakes"

AI doesn’t replace good structure it just exposes bad structure faster. Clean code makes change cheap. Messy code makes everything painful, even for AI. Garbage structure in, chaos out.

Are you setting it up so that you have some agents write code and others critique it/look for bugs? Playing them against each other sort of like GANs…

Xtr3m3hodl · 1d

Until they straight up have amnesia in between. Had an experience yesterday were Claude forgot about the task it was doing and had to remind it.

VariableFate · 1d

I've been repeatedly doing this. "Search for duplicated code, scattered logic, or inefficient methods. Suggest a list of targeted refactors ranked by risk/reward ratio.