AI should only run as fast as we can catch up

I have been a developer for twenty years now. For me to trust code, my want is to understand every single line. I learned long ago working on projects with a team that that becomes impossible for a single person on large projects. I learned to trust that someone understands the code and between blames and Slack I can almost always hunt that person down.

More and more often, while doing code review, I find I will not understand something and I will ask, and the "author" will clearly have no idea what it is doing either.

I find it quite troubling how little actual human thought is going into things. The AIs context window is not nearly large enough to fully understand the entire scope of any decently sized applications ecosystem. It just takes small peaks at bits and makes decisions based on a tiny slice of the world.

It's a powerful tool and as such needs to be guided with care.

Software becomes legacy very fast.

I have seen so many projects were people who understood all of it, are just gone. They moved, did something else etc.

As soon as this happens, you no longer have anyone 'getting it'. You have to handle so many people adding/changing very thin lines across all components and you can only hope that the original people had enough foresight adding enough unit tests for core decisions.

So i really don't mind AI here anymore.

Can humans though? There's a reason we don't just lump everything into one giant file and singleton class named DoIt(). Who hasn't come back around to some bit of code in a project and wondered what dumbass wrote this, only for the logs to tell you that it was you that wrote it, years ago. If AI is resulting in code that's more modular, in smaller digestible and understandable chunks, I'm not hearing that as a bad thing!

We might have to give up on trust and understanding in complex domains. To draw an analogy from another field, pharmaceutical researchers often don't understand the exact mechanism of action for drugs they develop. Biological systems are too complex and much of the basic research hasn't been done yet. So they rely on rigorous testing to verify that new drugs are safe and effective. It isn't a perfect system — sometimes drugs get recalled or have warnings added later — but works well enough.

Not sure why this is dead, but in nearly all of my consulting gigs sooner or later I ended up having to check on project/service that is effectively abandoned. Last time this morning. Luckily I had claude code and CLI tools to go through few dozen repos and millions LOC to find some obscure endpoints and data structures, since there wasn't even anyone to ask what to look for.

> I learned to trust that someone understands the code and between blames and Slack I can almost always hunt that person down.

Does your company not have many retirements, firings, or employees who quit to work elsewhere?