{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreigf4cixbxygmen2zuy7rb7nzpp4zyplxoeqp7yq2xho7b35od4dsa",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mix4i4obqwq2"
  },
  "path": "/t/would-this-concept-model-work/175056#post_3",
  "publishedAt": "2026-04-08T01:07:39.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "GitHub - Fury7425/bitDiffusion-a4.8 · GitHub"
  ],
  "textContent": "Thanks for the reply. Tbh I lied a little bit….\n\nI trained a 100m model on about 600m tokens(only pretraining) and it got a coherent output not the best in any means but i mean at that point. And also I do have somewhat of a cleanesk code. Used claude to write most of it but with my limited knowledge it does seem okay. This is the repo GitHub - Fury7425/bitDiffusion-a4.8 · GitHub and I asked claude to write readme for me based on the conversation that leads to this thing. If you want please try reading this report and like tell me how it would work or not work.\n\nAnd just ignore the lisence part. Idk why opus pug it there but I’m out of the daily limit to corrent it for the time being",
  "title": "Would this concept model work?"
}