{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreie3j6ra7auxvdo5muv6tyygazd5r2mcdfpigeesfpykcgv6ncbi2a",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mh7lyg2cznf2"
  },
  "path": "/t/how-far-before-simulation-isnt/170176#post_16",
  "publishedAt": "2026-03-16T20:54:29.000Z",
  "site": "https://discuss.huggingface.co",
  "textContent": "I haven’t actually distributed my persistence engine. I tried to distribute a demo of it months ago but no one was interested in it at all. So I took it upon myself to see how far I could go with it and developed an entirely new operating layer for AI using the persistence engine as the kernel. I am not really into the whole larp thing people get into with AI. I am more than happy to share screen shots, demonstrations and what not if you’re interested. It’s an Operating System just for AI which basically lets you swap and change models on the fly while sharing 1 source of truth with run trace and everything else already baked right into it. Is a complete control plain for AI with a swiss army knife of tools available for it to use. I also wired full OpenAI support straight into it. That’s ChatGPT 5.4 who replied in that screen shot. It can orchestrate the local models via tool calls and tell the local models what to do too. The automatic loop also makes all kinds of automation possible too. The Persistence Engine is the game changer. I also made it possible to digest any sized dataset directly into the filesystem which then makes it completely indexed immediately for the AI model to use. Also means everyone can import their entire chat historys from their online ai chat accounts into 1 session.\n\n\n\n\n\n",
  "title": "How far before simulation isn't?"
}