I've been procrastinating on designing extremely targeted AI agents to aid me with research with some context of my codebase.

I am starting to think this was a mistake. there is massive caveats and again, they need to be highly targeted, but I can now start seeing a glimpse of how they can be helpful.
there's a lot of research adjacent stuff that we spend a bunch of time on that there's no reason an AI agent can't take care off.