I'm curious how reviews happen for such huge PRs (120k lines). Do reviewers sit and go through all these changes over days?
If I had to guess, some humans skim things quickly for structural red flags, a bunch of LLMs do reviews based on various humans prompting to look for mistakes/ bugs, and then "tests pass == the code is good to merge".
The reviews don't happen...
alt Hacker News
If I had to guess, some humans skim things quickly for structural red flags, a bunch of LLMs do reviews based on various humans prompting to look for mistakes/ bugs, and then "tests pass == the code is good to merge".