How to improve AI coding/debugging in large codebases

Question

I mostly use Github Copilot(paid by the company). Are there any particular techniques that are helpful in debugging large codebases? So far I just have a MD file, with some general details about our codebase, and then I just ask it to search for a particular function/file and start from there. Are there other ideas that are particularly helpful for larger codebases, especially since this area seems to be changing rapidly?

kkarpkkarp · Accepted Answer

try cursor maybe. While I still try to provide as big as possible starting prompt when I try to introduce a new feature, I found for debugging it just works perfect. I almost never need to provide a big context.
Sometimes I get the whole wordpress site with ~20 plugins installed from customer, so it is huge codebase. But if customer tells me "i need to fix this button not sending this or that" that all what I copy to cursor and I have the fix on the first try almost every time.
I don't need to inform AI about anything more. Cursor figures it out itself: that it is wordpress, which plugins it has, which plugin is responsible for that button and how to fix it (mostly as filter/action to be added to the theme, sometimes creates custom tiny plugin)
The is also debug mode in cursor but I don't need it in that case. It is helpful when you are working on a new feature but at some point cursor stucks and can't produce what you ask. Then I switch to debug mode and wait when it fix it.