There’s a lot here but to pick a small point to argue, have you been brushing up on your assembly skills? Perhaps you keep in mind the internals of TLS or wonder about ECS in games? The stuff that needs to be cared about has just changed one level up again. What remains is what remained before too, the critical thinking that led to the post. The work case seems to be that people don’t care. I’d have a hard time believing that before AI that codebase was good. Even in my work with the gnarly non-AI legacy, monstrous things remain unacceptable, it just won’t be merged, AI or not.
Practical suggestion: try to influence the automated code review for better, an improved model, or harness, instructions to complain about what annoys you, more context etc. and even examples of bad recent work. But it seems in your workplace it would be ignored and merges would happen anyway. That too is not new, AI just amplifies, and the solution remains a new job, as before, mostly, and for however long we still have them.
I forgot to add one thing. You can still have fun too. While colleagues add the 30th one-off script to the tree, you can make a bot that critiques it and puts it on Slack. Is that reasonable? Doesn’t matter, make it fun and use some obscure language and learn while at it. That test suite you know would be amazing but never had time or mindshare to build? Side project it away, building up your AI tooling to the point you can just ask for it. Those N repos with broken makefiles? Same. Fun is still around, just changed.