Blog / code quality
How to keep an AI-built project from becoming a mess
A model will happily add code forever. Without a few habits, a project turns into files nobody understands, including you.
Commit after every working step
A commit is a save point. When the next change breaks things, you can go back in seconds instead of untangling the damage.
Keep a short project note
One file that says what the project does, how to run it and the main decisions. Paste it at the start of a new AI session so the model has context.
Ask it to explain before it edits
Have the model describe what it plans to change. Catching a bad plan costs a sentence. Catching a bad change costs an afternoon.
Delete what you do not use
Models leave dead code behind. Periodically ask what is unused and remove it. Less code means fewer places for bugs.
Name things clearly
Files and functions with plain names make future prompts easier, because you can point to them exactly.
Test the important path
- Write down the three things that must always work.
- Check them after every big change.
- Add an automated test for each when you can.
Log the decisions
When you choose a library or reverse a design, say why in your build log. It is the cheapest documentation there is.
Keep your own build log
Write numbered entries, follow other builders, and let people watch your idea become real.
Join vibFolio