{
  "author": {
    "name": "Dylan Fitzgerald",
    "url": "https://dylanfitzgerald.net/"
  },
  "blyg": "0.4",
  "changelog": [
    {
      "at": "2024-11-23T04:54:00Z",
      "note": "Imported from the Ctrl+Shift archive",
      "version": 1
    }
  ],
  "content_hash": "sha256:1cc6db2c595dd526c3b92ab20c7fdbb867916429c6092267dd1b50cd6b991565",
  "content_html": "\u003ch1\u003eA somewhat useful LLM framing: the sparring partner\u003c/h1\u003e\n\u003cfigure\u003e\u003cimg src=\"https://dylanfitzgerald.net/ox-hugo/daltoris_a_kung_fu_artist_ready_to_spar_against_a_robot_--v_6_3e790349-097b-470d-90cf-db5b60e42aac_3.jpeg\"\n\t\t\talt=\"A kung fu artist ready to spar against a robot\" width=\"50%\"\u003e\n\u003c/figure\u003e\n\n\u003cp\u003eIf you\u0026rsquo;ve tried to have an LLM write prose for you at some point\u0026ndash;particularly within the last year, since tightened safety mechanisms made it nigh-impossible to effect a more artistic voice\u0026ndash;you\u0026rsquo;ve probably been disappointed with its dry, stilted voice. The only available personalities appear to be \u0026ldquo;printed user manual\u0026rdquo; and \u0026ldquo;awkwardly trying to be brah\u0026rdquo;.\u003c/p\u003e\n\u003cp\u003eIt should go without saying that you shouldn\u0026rsquo;t publish this output. It\u0026rsquo;s awful, and often cringe to boot.\u003c/p\u003e\n\u003cp\u003eA more useful application is to ask the LLM to challenge and criticize your own writing. While its ability to generate new text with personality is weak, it knows well the rules of the game and can point out where you may have strayed. Ask: what about this misses the point? Where is my logic wrong? What seems redundant or extraneous? Convince me that my argument is wrong!\u003c/p\u003e\n\u003cp\u003eThe intent here isn\u0026rsquo;t to get the LLM to think and write for you. It\u0026rsquo;s to find weaknesses in your own writing and thinking.\u003c/p\u003e\n",
  "content_md": "# A somewhat useful LLM framing: the sparring partner\n\n{{\u003c figure src=\"/ox-hugo/daltoris_a_kung_fu_artist_ready_to_spar_against_a_robot_--v_6_3e790349-097b-470d-90cf-db5b60e42aac_3.jpeg\" alt=\"A kung fu artist ready to spar against a robot\" width=\"50%\" \u003e}}\n\nIf you've tried to have an LLM write prose for you at some point--particularly within the last year, since tightened safety mechanisms made it nigh-impossible to effect a more artistic voice--you've probably been disappointed with its dry, stilted voice. The only available personalities appear to be \"printed user manual\" and \"awkwardly trying to be brah\".\n\nIt should go without saying that you shouldn't publish this output. It's awful, and often cringe to boot.\n\nA more useful application is to ask the LLM to challenge and criticize your own writing. While its ability to generate new text with personality is weak, it knows well the rules of the game and can point out where you may have strayed. Ask: what about this misses the point? Where is my logic wrong? What seems redundant or extraneous? Convince me that my argument is wrong!\n\nThe intent here isn't to get the LLM to think and write for you. It's to find weaknesses in your own writing and thinking.",
  "created": "2024-11-23T04:54:00Z",
  "id": "0gbpss34vb36dxxb8mcg4dcaj3",
  "kind": "thread",
  "media": [],
  "origin": "https://dylanfitzgerald.net/blyg/",
  "page": "https://dylanfitzgerald.net/blog/the-sparring-partner/",
  "transclusions": [],
  "updated": "2024-11-23T04:54:00Z",
  "version": 1
}