ATProto Browser

Record data

{
  "uri": "at://did:plc:vbufq3xwt3233giwk4ulgvpr/app.bsky.feed.post/3loloardkuc2a",
  "cid": "bafyreidfthzdb5aq5julr6fhejoc5xkrxra6vjvgqyjrzjvg7woeaw7bvm",
  "value": {
    "text": "The error rate for newer “reasoning” models is going up, possibly because a lot of LLM generated content is ending up in the training set. \n\nThe idea that AI stuff gets better indefinitely has always been a convenient lie by its boosters. Has been since the 1950s.",
    "$type": "app.bsky.feed.post",
    "langs": [
      "en"
    ],
    "reply": {
      "root": {
        "cid": "bafyreibi2ixgzsxqsum4hoh6fwnuaxilqhabibuqegwbwaybrupyszfxpy",
        "uri": "at://did:plc:gcoiabsumehr3g6sbimjrqpx/app.bsky.feed.post/3lojlginjys2r"
      },
      "parent": {
        "cid": "bafyreigiokapcljxo6lxzhelnuuz5nfwdv52tvpqgyk2zfkygidf7uosue",
        "uri": "at://did:plc:xsgw6n6tvrcrgdiz5p5m64yq/app.bsky.feed.post/3lolij6yrl22r"
      }
    },
    "createdAt": "2025-05-07T15:04:41.053Z"
  }
}