While the product sounds mildly interesting, I see it as a major red flag you think it's ok for either a submitter or reviewer to not even read the code they are working with and ship thousand line diffs of LLM-generated code.
That's the lack of professionalism I give my random PoC personal projects where the only user I can break is myself - at work I am reading every line of every PR I submit or review, even if I used an LLM to assist writing the code.
I understand what you mean here but I'd consider this something that is configurable. I'd 100% have this product type as the default to require manual intervention/actions. Forcibly require someone to turn it off w/ some very explicit conversations. Maybe even provide some educational content about best practices.
Most of my previous companies required attaching a loom/screen recording of visual features cause the code really only communicates the logic. I've found that even for the PRs where you want to be super thorough and read every single line of code, watching the PR get tested brings you up to speed a lot faster.
its more of a comment on human nature. on average when we did monitoring at scale reviewers averaged around 150 lines of code actually reviewed. thats the normal, day to day human limit. human attention is woefully unable to scale to meet the demands of how much AI slop is getting pushed and this is a swing and helping solve this by valuing human attention spans. This isn't going to solve it all at once but we're working on it!
That's the lack of professionalism I give my random PoC personal projects where the only user I can break is myself - at work I am reading every line of every PR I submit or review, even if I used an LLM to assist writing the code.