I'm tired of reviewing one, two, three thousand line PRs because an AI agent "one shot the whole issue." Small PRs were never about ease of writing; they exist for the reviewer's benefit. AI is a boon but becoming a liability for maintainers. Arguments that "it won't work without the whole change" miss the point: small PRs provide digestible, reviewable, understandable pieces", not finished products. Comprehension time increases exponentially** with lines of code; taking exponentially more reviewer time doesn't spark joy.

Excessive comments are also unnecessary. Document functions with proper docstrings, but five lines explaining a variable named `is_logged_in` indicates poor naming—rename it instead. For AI maximalists saying "just use AI to grok it": you're wasting tokens re-ingesting AI-generated code. If you use a different model for review, why involve humans at all? Let AI split the PR first, then we'll review.

When React launched, we didn't accept larger PRs because "React is faster," so why now? Addendum: are you making giant PRs hoping I'll approve halfway through? Well played.