{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreigiuktz3cqyagdbyvm6cskiakdcapdkzfk7mix3lgwyzepfnqkm6e",
    "uri": "at://did:plc:pi6woz4d47bkuws673w2il2r/app.bsky.feed.post/3mlgdarzusw32"
  },
  "path": "/t/anti-llm-sentiment-considered-harmful/14008?page=5#post_101",
  "publishedAt": "2026-05-09T11:24:33.000Z",
  "site": "https://discourse.haskell.org",
  "textContent": "> I’m not sure why people are complaining about the tiresome effort of reviewing PRs when this is something that AI is unambiguously good at.\n\nHaving used AI for this I disagree strongly. It has strengths but it also has obvious flaws.\n\nYes it’s great at finding fiddly errors. And for certain changes I use such tools to screen for this kind of error once I think It’s done.\n\nBut for complex changes often the feedback can be quite misleading. With those tools not “understanding” or failing to consider certain invariants and things like that. Pointing out perfectly fine things as incorrect leading to a lot of wasted time sometimes.\n\nBut that’s not even the real danger imo. The real danger is being tempted into thinking the review by the AI is good enough and that resulting in no human review at all.",
  "title": "Anti-LLM Sentiment Considered Harmful"
}