How I used agents to study open-source code

Earlier this year i was studying some open source code. From projects like: Wild, Linux, binutils, MetaCall. I used this ASIDE.md (inspired by my friend Hossam) to help me, giving the LLM the oppurtiunity to teach me while i have to write the answers myself.

The main idea is that i investigate the project, and put my questions inside an ASIDE.md the LLM only answers my questions in chat. Then, I have to write the answers myself in the md file.

This pushed me into investigating code and understanding concepts, without the AI taking over, giving myself a room to breathe and check if i really understand what the LLM has written, if i don't i just keep asking, deep down in the rabbit hole.

Even further, if you have a good material you wanna study/discuss, you could just put it in another md file and keep asking the agent about it in ASIDE.md, so it now has the codebase as a reference so everything you discuss is evidence-based.

link to an example of ASIDE.md on a GitHub Gist: https://gist.github.com/fahdfady/6557295d131e49a703b0e4850f07eeb5