How do engineers internalize the codebase now? Do they just ask Claude and nod their head "because Claude said so"?

Yes, and osmosis

+Coneheads and Telepathy.

yup. and it's incredibly annoying, specially when a new joiner is writing a PR that a more experienced employee is reviewing.

i had to argue with someone, more than once, that "cursor is wrong about this assumption because i originally wrote this" because they just accepted whatever model they used as truth.

It's all gone if you don't take time for it, sure.

A good lithmus test I've noticed is "I want to do this thing the system should do", and if claude identifies a major refactor - well that's a sign the code needs a human touch first.

I'm finding it to be just the opposite - futzing around in the codebase is more effective than ever, it just looks different.

My futzing around is now at a higher level and involves enlisting the help of agents to dig through the code and logs and other sources to help solves problems.

Right now I am working an effort to reduce the amount of "build friction" across our code base. Even scoping this out would have taken a non-trivial amount of time and fixing it would never have been funded. With the help of AI tools, I can now attach this problem and make the lives of hundreds of engineers better.

> great way to find dead code, suboptimal structures, etc. With AI that's all gone.

it's not gone, it's literally a prompt away. I run "tech debt" loops all the time, ask it to simplify the code and architecture, remove left over code from previous refactors.

the quality of the resulting code is better than ever

> the quality

Measured by?

check for dead code, duplicated, SOLID, DRY, and KISS are now enforced on every push as part of our CICD, other agent does this of course, no one can check so many code, still better than before when we where trusting some energetic programmer would do