Comment by ReverseCold

> I can run and test the code at my discretion, whereas the AI model can't.

It sounds like you know what the problem with your AI workflow is? Have you tried using an agent? (sorry somewhat snarky but… come on)