{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreihpltdprbgdacmvx2ppenktvrktzqk2y33k2ljgvaevps52co2uby",
    "uri": "at://did:plc:cmylgpl5xczdixbceohfjpo7/app.bsky.feed.post/3mmjg4qg7fz62"
  },
  "path": "/@julesh/116623812317783961",
  "publishedAt": "2026-05-23T12:09:11.000Z",
  "site": "https://mathstodon.xyz",
  "textContent": "If I understand correctly, recurrent neural networks and transformers embody the 2 broad possibilities for sequence recognition: RNNs encode data in time, transformers encode data in spaceI wish I could know whether the cortex works fundamentally more like RNNs or more like transformers, or if it's the wrong question and the answer is a secret third option.",
  "title": "julesh (@julesh@mathstodon.xyz)"
}