Joined a new company recently as a Tech Lead for an experimental/R&D team.

We are moving at a breakneck pace—lots of POCs, rapid MVPs especially from huge team of juniors without review culture.

Because we’re leveraging Claude/LLMs heavily, the sheer volume of code being generated is insane. I'm regularly seeing PRs that touch 40+ files.

Tried chinkyfying - PR stacking, AI Reviews & explanations but still feels like using "Old World" processes for a modern, fast-paced environment.

Even with stacking, the overhead of context switching, GitHub UI limitations, and waiting for CI/CD on 10 tiny PRs is becoming the bottleneck.

If the AI can write 50 files of high-quality boilerplate and logic in 10 minutes, but the human-led review and merge process takes 2 days, the "AI advantage" starts to evaporate.

How are you all handling this?