I’ve been using gen AI tools like GitHub Copilot for a while now, and while they’re great for quick snippets, I’ve run into some issues. Copilot often misses the bigger picture of my projects, like recognizing patterns in my code or reusing functions from other modules. This means I end up copying and pasting code into ChatGPT a lot to get more complete answers, which isn't the most efficient workflow.
The problem is, ChatGPT tends to ALWAYS rewrite entire chunks of my codebase, sometimes changing function signature, breaking unit tests and causing more work. It’s clear that while AI is super helpful especially just to get a some initial code out, it’s not perfect yet for maintaining the broader context of an application especially as it increases in complexity.
I’ve been trying to balance using AI for initial coding and repetitive tasks while still doing the integration and refinement myself. It helps, but I’m curious how others are handling this. Are there any tips or tricks you’ve found to make AI more context-aware and less disruptive to existing code? Maybe I'm missing something entirely?
