{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreiewieb3jzc5p7kae5prpufc74fukh5uaohnfvh7qilfzi3yg4xniu",
    "uri": "at://did:plc:pi6woz4d47bkuws673w2il2r/app.bsky.feed.post/3mlja476fj5h2"
  },
  "path": "/t/anti-llm-sentiment-considered-harmful/14008?page=6#post_106",
  "publishedAt": "2026-05-10T16:04:25.000Z",
  "site": "https://discourse.haskell.org",
  "textContent": "I don’t think that is what’s meant here though. The LLM can help with the boilerplate that is necessary. The boilerplate that needs to happen.\n\nNow AI may have a tendency to generate more boilerplate than necessary, but imo that’s precisely where the engineer is to iterate with it. Besides writing boilerplate, AI is also excellent at refactoring. With the engineer’s guidance to make a good abstraction, the AI can help apply it at scale. This is why I genuinely believe that these tools in the hands of a strong engineer will actually produce _higher_ quality code, given the same time to work on it.\n\nWith AI, you typically get a working, but “not great” solution very fast. That gives you a lot of time left to iterate, build and apply abstractions, refactor, maybe even try different paths of refactors. If it’s easy to try something out, it’s also very cheap to throw it out, take a step back and walk in a different direction. I’ve done this a couple of times, where I went down the path of a refactor, saw the consequences halfway, regretted it, and just went back to do something else instead. The code was always better for it, and very little time was wasted on the wrong path.\n\nFor reference, I strongly distinguish “vibe coding” from “AI Augmented Engineering”. With the former, people barely look at the code at all, and work exclusively by talking to the bot. With the latter, proper software engineering skills are used to ensure a high standard of quality. The former I reject as bad practice, and I _think_ everyone in this thread (on both sides) would agree. The latter is something where I do see a lot of value.",
  "title": "Anti-LLM Sentiment Considered Harmful"
}