oh, i can already do this kind of stuff but it doesn't inter...

mleku

npub1fjqqy4a93z5zsjwsfxqhc2764kvykfdyttvldkkkdera8dr78vhsmmleku

hex

ec6c7658ac20f438c1ea030c617c87ee17dc5d6ef30aa78e28d8e285da9f98d0

nevent

nevent1qqswcmrktzkzpapcc84qxrrp0jr7u97ut4h0xz483c5d3c59m20e35qprpmhxue69uhhyetvv9ujuem4d36kwatvw5hx6mm9qgsyeqqz27jc32pgf8gynqtu90d2mxztykj94k0kmttxu37nk3lrktc89cv4e

Kind-1 (TextNote)

2025-12-03T22:05:46Z

↳ Reply to halalmoney (npub1vdaz78d96t7pjakpfmqqlr4wgq26l6vmkf25782n05qgdgtfmk8s297rjz)

I meant all that self hosted llm stuff! But happy you found it useful in another way.

oh, i can already do this kind of stuff but it doesn't interest me that much. i'm not really searching for any new things. me and books and me and girlfriends i am the same with it... just sorta feel like when it's The Apocalypse of Yajnavalkya time, that Mark Jeftovich's writer will post me a tantalizing excerpt. well, that was two years ago, now i expect it will appear on nostr, and actually, a lot of interesting stuff has come my way through nostr, in amongst the noise.

nah, i'm most interested in the coding/planning/reasoning stuff. that's why i don't use my GPU for much other than games, it can only run a 20b model which yeah can do this kind of summary/analysis stuff.

no, i want to train an LLM to be a Rob Pike level go programmer. and Leslie Lamport level cryptographer. and hopefully, Richard Feynman level physicist.

i'm pretty sure the techniques required don't exist yet and that HRM stuff gave me a lot of ideas about how to use the inherent commutability of very wide variations of an expression of an algorithm in code, that can be shuffled by taking time dependencies into account. this kind of preparation of training data is known to greatly accelerate training, and the quality of reasoning in the field the data set pertains to.

Raw JSON

{
  "kind": 1,
  "id": "ec6c7658ac20f438c1ea030c617c87ee17dc5d6ef30aa78e28d8e285da9f98d0",
  "pubkey": "4c800257a588a82849d049817c2bdaad984b25a45ad9f6dad66e47d3b47e3b2f",
  "created_at": 1764799546,
  "tags": [
    [
      "e",
      "9be783927c98c2bef9d703fc6a06dbbe78a19f31a2a3bde7404774f48d4ff58d",
      "wss://ditto.pub/relay",
      "root",
      "55f573b651eff351db57b0601d23022d8c532f9825db10a5733ebf39be4aa21b"
    ],
    [
      "e",
      "6afc14bbfba74d99a42a6a90204f90d77d3542c81a63562708298cc40e2ca894",
      "wss://theforest.nostr1.com/",
      "reply",
      "637a2f1da5d2fc1976c14ec00f8eae4015afe99bb2554f1d537d0086a169dd8f"
    ],
    [
      "p",
      "637a2f1da5d2fc1976c14ec00f8eae4015afe99bb2554f1d537d0086a169dd8f"
    ],
    [
      "p",
      "55f573b651eff351db57b0601d23022d8c532f9825db10a5733ebf39be4aa21b"
    ],
    [
      "p",
      "9a21569255d0a3a9e75f1de2e4c883c9be2e5615887f22b2ecf6b1813bcd587d"
    ],
    [
      "client",
      "jumble"
    ],
    [
      "alt",
      "This event was published by https://jumble.imwald.eu."
    ]
  ],
  "content": "oh, i can already do this kind of stuff but it doesn't interest me that much. i'm not really searching for any new things. me and books and me and girlfriends i am the same with it... just sorta feel like when it's The Apocalypse of Yajnavalkya time, that Mark Jeftovich's writer will post me a tantalizing excerpt. well, that was two years ago, now i expect it will appear on nostr, and actually, a lot of interesting stuff has come my way through nostr, in amongst the noise.\n\nnah, i'm most interested in the coding/planning/reasoning stuff. that's why i don't use my GPU for much other than games, it can only run a 20b model which yeah can do this kind of summary/analysis stuff.\n\nno, i want to train an LLM to be a Rob Pike level go programmer. and Leslie Lamport level cryptographer. and hopefully, Richard Feynman level physicist.\n\ni'm pretty sure the techniques required don't exist yet and that HRM stuff gave me a lot of ideas about how to use the inherent commutability of very wide variations of an expression of an algorithm in code, that can be shuffled by taking time dependencies into account. this kind of preparation of training data is known to greatly accelerate training, and the quality of reasoning in the field the data set pertains to.",
  "sig": "b53de3fd0dad8ae31862fed69fe630af258db29835df715a0a8cc003c757b8ce7be0154159c5f0ea6e37d1288271a407f1de961f482bf6b8d57940db553e1a48"
}