There's a funny thing about AI-generated code...
AI is remarkably good at producing a convincing proof of concept.
But as you add features and make adjustments, something interesting happens.
The codebase grows bloated, the architecture becomes convoluted, and the application…
Commit
d90a697
Well, at least AI also makes it easier to throw that away and re-implement. But it's on you to provide proper guidance.
True. But then, how well do you understand the code you are deploying should an issue arise?
You can set up a system for scanning, identifying and reviewing the key paths/decision trees that should be reviewed before you deploy.
The definition of "key" here is different for everyone and the projects they're working on. And the solution is based on that context.
I don't think one can keep up with the required velocity level if you review every line of code.
TLDR - Define sensible filters and rules - create a tool to enforce and apply them. It should follow your project needs.
commit d90a697