Login
You're viewing the mastodon.coffee public feed.
  • Daisy Leighdaisy@cloudisland.nz
    Aug 9, 2026, 4:19 AM

    This isn’t an original idea but rather a distillation of the ideas of others in a way that made it click for me, but:

    AI slop code review is so much more draining because it is a fundamentally different task; your job goes from verifying and sense checking someone else’s human understanding to, oftentimes, developing the human understanding that the “author” has not bothered to undertake at a more-than-perfunctory level. Which is exhausting in itself and ALSO violates the social contract of code review.

    💬 1🔄 9⭐ 9

Replies

  • Joshua Leungaligorith@mastodon.nz
    Aug 13, 2026, 11:57 PM

    @daisy
    The two things that have not gone down well for our team reviewing slop yeeted over from the other side:

    1) Encountering all the oodles of auto-generated tests and docs, that are copious in verbiage and seemingly correct (in that they do superficially seem to do justifiably valid stuff), yet just "feels wrong" at the end of the day...

    That stuff is just draining to have to go through as a reviewer. You also wonder if the author even bothered to check themselves after the first few times coming across it. Then you start wondering if letting this through sets a precedent that then gets used to mean that the tests are now all unmodifiable without having to use the slop tools to regenerate them later, as they are now unmaintanable otherwise...

    2) More demoralising is giving a standard well intentioned comment, only for the author to seemingly have their slop machine generate a reply directly (along with fixes)... and then, on their very next issue, they commit the same infraction a dozen more times!

    At that point, you get a very dirty feeling of having just ending up having prompted someone else's AI for them, by proxy, while the human you thought you were working with has just abdicated responsibility...

    💬 0🔄 0⭐ 0