Curt Doolittle

Source: Twitter X

  • (Diary) Cat making a racket outside. Young Male. Very fluffy white-orange. Why?

    (Diary)
    Cat making a racket outside. Young Male. Very fluffy white-orange. Why? Wants to bring me dead things as a present. (sigh). It’s not even my cat. I just let him in when he wants attention. And see? I get presents. (sigh)
    I don’t want dead presents. 😉

    Source date (UTC): 2025-09-10 17:38:28 UTC

    Original post: https://twitter.com/i/web/status/1965832258516070711

  • Our Training Data How we work: Research > Reduce to Book Form (a system) > Feed

    Our Training Data

    How we work: Research > Reduce to Book Form (a system) > Feed to AI > Get Training Plan > upload our system prompt (Prompt_Protocols) > Pick the next module > Ask the AI to produce the training examples for that module > in our case that’s socratic And

    When we build a training plan for one of your books, each module consists of a range of assertions that includes:

    1. Canonical Assertions –
      These are the core, necessary, and sufficient statements of fact, principle, or law in your system. They are crafted for       maximum precision and serve as the standard reference points for truth, decidability, and operational rigor. They carry the full weight of the framework and must pass the highest bar for testifiability.
    2. Adversarial Assertions –
      These intentionally introduce       edge cases, counterexamples, or potential failure modes. They test whether the system can withstand criticism, falsification attempts, and hostile interpretation. Adversarial assertions ensure the framework isn’t just self-consistent but also resistant to parasitism, ambiguity, or strategic misrepresentation.
    3. Exploratory or Speculative Assertions (if included) –
      These identify       open questions, conjectures, or contingent hypotheses that extend beyond current proofs but remain operationally plausible. They guide future research or refinement without diluting the canonical set.
    4. Didactic Assertions (optional but often useful) –
      These restate canonical ideas in       simplified, pedagogical, or narrative form for teaching purposes, ensuring accessibility while preserving precision.
    So by using PROTOCOLS, training examples in our OUTPUT_CONTRACT in analytic form, and then using the SOCRATIC form we explicitly add the Didactic Assertions by using the socratic form. Sort of ‘belt and suspenders’.

    So we use all four set so training assertions to achieve both the accessible interface and the deep interface customers need.

    Source date (UTC): 2025-09-10 14:53:08 UTC

    Original post: https://x.com/i/articles/1965790649195872617

  • RUMORS OF AI USE DECLINE IN BIZ ARE NONSENSE (a) it’s a tiny 2% decline on a sma

    RUMORS OF AI USE DECLINE IN BIZ ARE NONSENSE
    (a) it’s a tiny 2% decline on a small number in the first place (meaningless)
    (b) It’s because the initial interest level was over-reported due to status signal fears.
    (c) It looks like they found the source of hallucination (and it was obvious).
    (d) We (NLI- Runcible) have the solution to ‘truthfulness’ and ‘data curation’ – so the quality problem is solvable.
    (e) Their problem is they mixed the data with alignment. Whereas to know the truth you must discover it first, and then align for the government, culture, or user as required or preferred afterward. This separates the logical reasoning from the delivery of the message.

    The primary issues I have found is that they raise the temperature to get more ideation and less precision, and that the alignment bias is integrated into the foundation model, so that not only is the data un-curated, not only is it contain falsehoods, not only is it postwar normatively biased, not only is it feminine sensitivity biased, but together it’s very difficult to reason on one hand, and very difficult to constrain it from bias in the second.

    Source date (UTC): 2025-09-10 14:29:10 UTC

    Original post: https://twitter.com/i/web/status/1965784619544141968

  • In my case I have apple hardware (and a lot of it) so I like that everything int

    In my case I have apple hardware (and a lot of it) so I like that everything integrates in every way possible.

    Source date (UTC): 2025-09-09 22:30:07 UTC

    Original post: https://twitter.com/i/web/status/1965543268172923005

  • “Runcible: Enterprises aren’t blocked by AI that can’t write — they’re blocked b

    —“Runcible: Enterprises aren’t blocked by AI that can’t write — they’re blocked by AI they can’t trust.”— Brad Werrell

    Source date (UTC): 2025-09-09 22:20:52 UTC

    Original post: https://twitter.com/i/web/status/1965540937796661729

  • (New IPhone Release?) I bought Windows CE when it first came out, then the nokia

    (New IPhone Release?)
    I bought Windows CE when it first came out, then the nokias that used it.
    I bought Iphone 2 and kept it through 4.
    I kept the 4 through 6.
    I kept the 6 through 12.
    I cant think of a reason to upgrade from the 12 unless … something like a week of battery life?

    Source date (UTC): 2025-09-09 18:29:12 UTC

    Original post: https://twitter.com/i/web/status/1965482636530602236

  • (There must be an opportunity to tease you in there somewhere by I can’t find it

    (There must be an opportunity to tease you in there somewhere by I can’t find it. So I’ll send hugs instead. lol)

    Source date (UTC): 2025-09-09 18:14:52 UTC

    Original post: https://twitter.com/i/web/status/1965479031123705979

  • (Diary) You know what I love most about working for myself? I get to take a nap

    (Diary)
    You know what I love most about working for myself? I get to take a nap whenever I want. 😉 Like, I get up at 5 or 6am, work a bit, have a few phone calls, go out to brunch, and come home on a grey day in the early fall, and just decide that a nap is the best possible use of my time. 😉
    Glorious. 😉

    Source date (UTC): 2025-09-09 18:10:49 UTC

    Original post: https://twitter.com/i/web/status/1965478010808926577

  • Jack. You are correct of course. But then again that is the function of memory.

    Jack. You are correct of course. But then again that is the function of memory. To compress information so that it can be retrieved and acted upon faster than the environment can change so that we can defeat time, and in doing so capture its calories. 😉

    Thought for you: compression and reduction. Ie: reducibility. Reduction permits compression. Mathematical, algorithmic, operational, formal language, ordinary language all reduce to compress.

    Source date (UTC): 2025-09-09 17:39:33 UTC

    Original post: https://twitter.com/i/web/status/1965470143641014646

  • UPDATE: We cannot tune (alter the weights) a custom gpt like we are using to dem

    UPDATE:
    We cannot tune (alter the weights) a custom gpt like we are using to demo CurtGPT. It’s limited to prompts and RAG (Files).

    So we will take it as far as we can today and tomorrow.

    After that we will move to 4o-mini, which we can use prompts, RAG, and training.

    Our training uses both Socratic and analytic examples for every item. So we train both conversational and judicial models at the same time.

    But again we are training from our internally consistent and commensurable volumes not from arbitrary incommensurable datasets that are the industry standard.

    Our goal is a realistic demo. And the custom GPT does a fair job. But we want to demonstrate the full effect – so we need to train another model. And no other ai is even close to as capable.

    Source date (UTC): 2025-09-09 17:33:35 UTC

    Original post: https://twitter.com/i/web/status/1965468642503737852