{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreicnugpbop4gebnlnvcp7upgoly7bk65cwvvvzvspd7wteuetj3fza",
    "uri": "at://did:plc:pi6woz4d47bkuws673w2il2r/app.bsky.feed.post/3mhbnsiyaiup2"
  },
  "path": "/t/yet-another-opinion-on-llms-hasufells-blog/13775?page=3#post_44",
  "publishedAt": "2026-03-17T10:43:18.000Z",
  "site": "https://discourse.haskell.org",
  "tags": [
    "Gell-Man Amnesia"
  ],
  "textContent": "Kristian:\n\n> The one thing that sticks with me is how well I think the LLM performs always seems to correlate with my knowledge of the problem space. The better I understand the domain and technology that is being used the worse the LLM is. Maybe the things I understand well are things LLMs are bad at? I find that hard to believe though.\n\nYou basically describe Gell-Man Amnesia but for AI capabilities. Something that happens for all of use to one degree or another I believe.",
  "title": "Yet another opinion on LLMs · Hasufell's blog"
}