{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreihablwndhh6cfybygquuw5mzim2g4kqqt4iobmxw4wp6vm46b2bpm",
    "uri": "at://did:plc:pi6woz4d47bkuws673w2il2r/app.bsky.feed.post/3mlyxci3actj2"
  },
  "path": "/t/anti-llm-sentiment-considered-harmful/14008?page=6#post_123",
  "publishedAt": "2026-05-16T20:34:25.000Z",
  "site": "https://discourse.haskell.org",
  "textContent": "I’ve been plowing through some boilerplate and did notice the constant nag in the background: “this is absolutely pointless inexcusable bullshit, nobody should be forced to do this manually”.\n\nThat reminded me of the procrastination equation (motivation = value * expectancy / impulsiveness / delay). AI tools cut delay and raise expectancy (assisted motivation gains). But over time the impulsiveness goes up (unassisted motivation loss). That means the tolerance for bullshit tasks drops. What was previously palatable with assistance now makes one riot. And for good reason!\n\nThe “laziness” here is indicative of the sad state of our tools, our libraries, our practice. One may say “writing Haskell brings me joy, I don’t see the point of delegating the joy to AI”. But that may not be true for everyone. And if someone prefers working with an ecosystem proxied by an AI that’s an alarm ringing.\n\nAnd this goes deeper, for the AIs also being lazy! And if the ecosystem makes slop easier/mentally cheaper than quality product - slop is that would dominate it.\n\nIf you’re the author of some library, I suggest testing some mid-tier AI to make a product with it. Spot where it takes shortcuts over “expected” usage. See what blunders it would make. Then improve around that. Your human users will also benefit. Ask AI to extend it in some non-trivial way, spot the hacks, improve until it does the right thing. Your contributors will also benefit.\n\n… and so I continued to plow through that boilerplate, barely holding it, for the sole reason of that activity was to measure the amount of anguish the library inflicts on its users. That amount is now lower, but still unacceptably high.",
  "title": "Anti-LLM Sentiment Considered Harmful"
}