{
  "blyg": "0.3",
  "id": "0vn788h3z7ejvpxgyhvqwhask0",
  "kind": "fragment",
  "origin": "https://www.msweet.net/notes/",
  "page": "668-maroon-snipes",
  "author": {
    "name": "Matthew McDowell-Sweet",
    "url": "https://www.msweet.net/"
  },
  "created": "2026-08-17T15:55:42Z",
  "updated": "2026-08-17T15:55:42Z",
  "version": 1,
  "content_md": "Thermodynamic computing might turn out to be the barefoot shoe of the computing world; demonstrably better but really slow adoption.\n\n> Guillaume Verdon joins MTS to discuss scaling thermodynamic computing, replacing classical bit architecture with pbits, and building software frameworks for stochastic programming. He outlines the trade-offs between numerical precision and power efficiency when running AI models on low-power chips. Additionally, he shares how thermodynamic systems can co-exist with GPUs to tackle complex frontier problems.\n\n[Classical Computing Cant Scale AI | Guillaume Verdon - YouTube ↗](https://youtu.be/J0jL8y_dDc8)",
  "content_html": "<p>Thermodynamic computing might turn out to be the barefoot shoe of the computing world; demonstrably better but really slow adoption.</p>\n<blockquote><p>Guillaume Verdon joins MTS to discuss scaling thermodynamic computing, replacing classical bit architecture with pbits, and building software frameworks for stochastic programming. He outlines the trade-offs between numerical precision and power efficiency when running AI models on low-power chips. Additionally, he shares how thermodynamic systems can co-exist with GPUs to tackle complex frontier problems.</p></blockquote>\n<p class=\"source\"><a href=\"https://youtu.be/J0jL8y_dDc8\">Classical Computing Cant Scale AI | Guillaume Verdon - YouTube ↗</a></p>",
  "content_hash": "sha256:42d37e38d885fdeb36184288aa55b5fc693ef3817c991f32c175fe8a610e6926",
  "media": [],
  "changelog": [
    {
      "version": 1,
      "at": "2026-08-17T15:55:42Z",
      "note": null
    }
  ]
}