I wrote a quick script to feed chatGPT lots of .h and .cpp files from a codebase, with the goal of getting chatGPT to read a few JSON files full of this code and provide high quality snippets.
The issue I’m running into is that it’s performing remarkably worse at writing code with this large quantity of data as an input.
Has anyone else been playing with feeding a codebase to chatGPT to get a smart AI helper?
It seems to ‘‘forget’’ everything it read in previous comments as soon as I start replying to it.