{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreihmvszgrstpuc22jl7vq5nu4vgi6uf5cgis4acqjqfnupmisllrcm",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mhcli672zid2"
  },
  "path": "/t/can-llms-be-computers-embedding-a-vm-inside-an-llm/1376724#post_7",
  "publishedAt": "2026-03-18T03:34:05.000Z",
  "site": "https://community.openai.com",
  "textContent": "So is the gain actually in efficiency, or just in not serializing the reasoning into tokens?\n\nFrom what I understand, this approach avoids emitting reasoning tokens, and instead performs the reasoning internally in the model’s state. I presume with similar underlying compute?",
  "title": "Can LLMs Be Computers - Embedding a VM inside an LLM"
}