I haven't followed the space with too much effort, currently I am only making use of ChatGPT to explain code snippets or ask questions about the Linux kernel source.

But by now, I bet there is already some better solution out there specific for the task of explaining large codebases, would love to hear what everyone here has tried out.

I do not understand the limits of LLMs, is it possible to have a model interpret a codebase it has not been trained on? How would one provide the source to it?