{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreif54chysk3agpevaenorfoxuxfto5ajvqovns66ej6wavtbn62sne",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mkvqv3zof7i2"
  },
  "path": "/t/rth-lm-25b-fractal-tcn-language-model-with-no-self-attention-looking-for-loi-collaborators-for-sprind/175720#post_1",
  "publishedAt": "2026-05-02T21:40:09.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "RthItalia/Rth-lm-25b · Hugging Face",
    "GitHub - rthgit/ZetaGrid · GitHub",
    "https://doi.org/10.6084/m9.figshare.31376560"
  ],
  "textContent": "Hi HF community,\n\nI built RTH-LM a 25B parameter language model that replaces\nself-attention entirely with a Fractal Gated Causal TCN architecture.\n\nThe core idea is to separate the system into a frozen reusable core\n(Genome, ~7GB) and modular trainable behaviors (Souls, ~300MB).\n\nSame Genome, different Souls: language and code generation are already\nworking without retraining the base model.\n\nLinks:\n\n  * HF: RthItalia/Rth-lm-25b · Hugging Face\n  * GitHub: GitHub - rthgit/ZetaGrid · GitHub\n  * Paper: https://doi.org/10.6084/m9.figshare.31376560\n\n\n\nI’m applying to SPRIND Next Frontier AI, a European funding challenge\nfor AI systems beyond the current Transformer paradigm. For the application, I’m looking for 2–3 European ML engineers or researchers willing to sign a Letter of Intent.\n\nThis is not a work commitment at this stage. It is simply a one-page\nstatement saying that, if the project is funded, you would be interested\nin joining the team. If funded, collaborators would have a real role in the project.\n\nHappy to answer technical questions here before anyone decides. Particularly interested in hearing from anyone working on non-transformer architectures, efficient inference, or alternative scaling approaches.",
  "title": "RTH-LM: 25B Fractal TCN language model with no self-attention, looking for LoI collaborators for SPRIND"
}