Curt Doolittle

Category: AI, Computation, and Technology

  • (Runcible) RE: LLM R&D TL/DR; Yeah, well, LLMs are really pretty dumb and I’ve s

    (Runcible) RE: LLM R&D
    TL/DR;
    Yeah, well, LLMs are really pretty dumb and I’ve sought and hit the limit of their abilities. And it’s been an exasperating journey of dashed hopes. 😉 (well, mostly).

    Problem Space:
    We produce:
    a) the runcible governance layer which is my epistemology applied to computability. It is extraordinary. It overcomes the problems of LLMs but not the limitations. It’s a wrapper around the LLM that binds it to limited pathways through the latent space.

    b) control over this layer has forced us to create and modify one of the open source models. We had hoped to ‘help’ openAI, but they are a bit ‘unfocused’ as a company and getting access to discuss at a high enough level amidst their many pressures is almost impossible on one hand and more costly than doing it ourselves on the others. So in effect we have a router llm, and a set of target LLMs, functioning as a mixture of experts. But realistically it’s just cost control.

    c) We also produce Oversing, which is a mobile, tablet, desktop universal application platform for running organizations of any sale, but basically it’s a host for contexts that you use to collaborate with one another and an LLM.

    Result:
    This is a crazy project of enormous scope but it’s the end point of what people need going forward. Cooperation at scale.

    No one has done this yet. It’s partly because they don’t know how to, and it’s partly because it’s hard. 😉

    Context:
    While I have built many tech companies, and consulted for the fortune 400, and made the Inc500 three times, I work in many disciplines, and my fundamental skills are epistemology, especially in high dimensional closure domains such as cognitive science, language, economics law, and evolution.
    The simple version is reducible to the fact that I don’t make mistakes that are common in the STEM fields by trying to reduce to mathematics or algorithm, that which is not reducible further than operational prose.

    So this is why I can solve this problem for LLMs.
    It’s just … let’s say… a little exasperating – just like every previous tech …until we understood it. 😉

    CD

    Source date (UTC): 2026-02-10 21:46:29 UTC

    Original post: https://twitter.com/i/web/status/2021340016082034787

  • IMO: 1) They are exceptional synthetic search engines. In other words – pattern

    IMO:
    1) They are exceptional synthetic search engines. In other words – pattern matchers.
    2) They imitate the human language facility. In that sense they are amazing. But they don’t imitate the neocortical spatial faculty, (this is LeCunn’s argument) or the hippocampal episodic memory formation, or the frontal cortex recursive wayfinding (testing an idea). They are bad reasoners because reasoning requires reduction to episodic steps, and wayfinding by recursion. And recursion is expensive.
    3) Our company’s governance layer( a wrapper around the LLM) can however determine whether a claim or assertion is true/false, ethical/not, possible/no, warrantable/not, liability-producing/not. But it does so by breaking down the problem and recursively testing each step.
    4) IMO leCunn is only half right in that we need a world model, but he is wrong, in that linguistic reduction isn’t a necessary property. It’s that you need an LLM (sematic store) for hypothesis generation (auto-association), the equivalent of a router (prefrontal cortex) to manage the ‘reasoning’ process (wayfinding) and to maintain states (episodes), a spatial model to test operational possibility, the llm linguistic model as the input output protocol.
    This means we’re just early and very demanding of one revolutionary insight, which is the attention model that mimic the human language facility, and the N-Dimensional sematic manifold as memory.

    CD
    Runcible
    NLI

    Source date (UTC): 2026-02-10 19:35:56 UTC

    Original post: https://twitter.com/i/web/status/2021307165441732671

  • IMO A) There are a limited number of agent patterns. B) very few agents are of a

    IMO

    A) There are a limited number of agent patterns.
    B) very few agents are of any meaningful value even to power users. Everything I have seen could have been easily produced by existing automation scripting or basic coding.
    C) Those few useful agents will rapidly be commoditized
    D) it’s not the agents that matter it’s the framework they operate within.
    E) Agents are ripe for every kind of hacking and abuse that has existed for the past forty years.
    F) There is every chance that enterprises will prohibit them unless each is approved.
    G) General insight that they increase noise and not signal is hard to argue with.

    I’ve been through every tech wave since the early seventies and it’s always the same game.

    Source date (UTC): 2026-02-10 19:21:10 UTC

    Original post: https://twitter.com/i/web/status/2021303445916115330

  • Just to confirm your point… At Runcible we produce a governance layer with the

    Just to confirm your point…
    At Runcible we produce a governance layer with the equivalent of your ‘stop’ conditions. It’s much more than that.
    It’s epistemic rigor for machines, and only LLMs have the informational density in semantic form to produce it.

    Source date (UTC): 2026-02-09 20:48:25 UTC

    Original post: https://twitter.com/i/web/status/2020963015647166843

  • “I wonder if the programming AI could be tricked into seeing NL as code and ther

    –“I wonder if the programming AI could be tricked into seeing NL as code and therefore applying it more strictly.”–
    @NoahRevoy
    NLI, Runcible

    Well I mean, it does – that’s why it works. Operational prose is just ‘code’ for human action at human scale in the existential reality we must navigate. That’s why we’re so strict about ‘enumeration, serialization, operationalization, and disambiguation into an identity”, and why we produce a dictionary, and dictionary terms on a dimension producing natural indexing and measurment – so that langauge becomes code.

    The AI’s (or at least be better ones) understand this and why we’re doing it. That’s why they can render the output that they do.

    Our problem (really) is that while we have created the language and the compiler, the present LLMs (operating system) are having as much problem running our ‘program’ with current memory limitations as did my original work in the 1980s using semantic indexes (tokens), possible actions (actions), and episodic memories (contexts) for predicting optimum choices (outcomes).

    I couldn’t do it (well) in assembler back then because of memory limits, and I’m having a heck of a time with 256K context windows doing it with LLMs today. I ran into the same problem building the first serious legal AI. Semantic depth is a memory burden because it’s a relational density burden because in turn, the information is stored in terms that are relationally dense.

    Whereas the human brain does it all in a massively parallel hierarchy, we have to produce domain, customer, individual protocols, then put them through our epistemic protocols to determine if they’re true.
    We could parallelize some of the epistemic protocols but again, that’s a cost.
    If we were to continue to use OpenaI for a hard question we could burn $2 per analysis and more for a certification. Whereas for most people with most questions our ChatGPT Custom GPT will do a better job already than any other LLM.
    Fundamentally any of these LLMs without compartmentalization produce drift just like people with ADD produce drift.

    So yes, it’s code. And the LLMs are operating systems that can run semantic code. But they were trained to favor normativity instead of truth, so until we can audit an entire 1T+ parameter LLM (which costs $$$$$!) we won’t have an operating system to run our ‘program’ on that doesn’t basically insert error.

    Cheers 😉

    Source date (UTC): 2026-02-06 23:11:42 UTC

    Original post: https://twitter.com/i/web/status/2019911909915721836

  • AFAIK if you produce the context the AI will pursue the context. if you train an

    AFAIK if you produce the context the AI will pursue the context. if you train an AI on normativity (correlation). you will get normativity. Human normativity is full of falsehoods, deceptions, propaganda, bias, and excuses for criminal activity.

    Source date (UTC): 2026-02-06 06:22:33 UTC

    Original post: https://twitter.com/i/web/status/2019657951175401716

  • @leecronin The amount of work we’ve put into de-glossing the AI’s so that we get

    @leecronin

    The amount of work we’ve put into de-glossing the AI’s so that we get truthful answers out of them is absurd.

    AI companies sought to produce assistants for the general public and have effectively produced them for programmers.

    They did not design them for testimony.
    (That’s what we did.)

    Prof: Thanks for your efforts in public awareness of science.

    Source date (UTC): 2026-02-06 06:19:49 UTC

    Original post: https://twitter.com/i/web/status/2019657261480833444

  • The Impact of AI on Artists In Historical Context: There’s Nothing New Here. Shi

    The Impact of AI on Artists In Historical Context: There’s Nothing New Here.

    Shifts in art always track four things: tech, money, housing, and the economy. All industries go through the same sequence of revolutions whether technology that enables more people at lower cost to do same or better work, or whether it allows more people at lower cost to consume art products.
    Personal Anecdote:
    While it might seem surprising for a serial technology founder, philosopher and social scientist, my undergraduate training is not in economics or law, but in fine art, art history, and theory. One of the more effete degrees one can obtain. And my first serious work of analytic philosophy was in the valuation of art. In fact, the general method I used for measurement of the subjective is the foundation of all the work in high-closure domains that I’ve used in my intellectual work products.
    An Example
    I remember in the early 80s having a debate with a then rather famous artist not too happy with my questions (as usual). He expressed the exasperation that at different times different peoples invested in art. And he gave examples. Of course, I explained that all the examples he used were the product of empires concentrating wealth sufficiently that they could virtue signal by hiring artists whose works otherwise would be too costly to produce. He was a usual mid twentieth product of the left just as most people in the entertainment business are today. He didn’t like my explanation – of course.

    Soon after I was a young but senior exec at the worlds largest art supplier. Hundreds of locations. But the day I saw desktop pubishing combined with digital typesetting I told the CEO we had to sell and get out of the business as fast as possible.

    A THOUSAND YEARS OF ARTISTIC JOB TRANSFORMATION
    Pre-1400s: Illuminated manuscripts. Monks hand-copied books, added gold leaf, intricate miniatures, pigments—each a unique luxury object for nobility and church. Gutenberg’s movable type press (c. 1440s) + woodcut illustrations mass-produced texts and images. Scribes and illuminators lost work; craft shifted to rare high-end editions or died out. Monastic scriptoria faced existential threat; many scribes retired or adapted to new roles like proofing printed books. Woodcuts democratized visuals in Northern Europe—DĂŒrer mastered them for wide sale, undercutting one-off hand drawings.
    Renaissance reproductive engravings (late 1400s–1500s). After DĂŒrer and others mastered intaglio engraving, prints reproduced famous paintings and drawings at scale. High-investment frescoes and panel paintings became replicable merch for the middle class. Painters saw commissions drop for exact copies; many adapted by producing originals for elite patrons while engravers handled mass distribution.
    1500’s With the restoration of commerce after muslims closed byzantium’s ports and the age of sail, we saw the dutch art flourish in order to satisfy the demands of the new middle class wealth. Oil paints got cheap and portable, so suddenly every merchant could commission Madonnas for the mantle. Dutch school, Flemish school, all the guys. The same week the Vatican flooded the market.
    Soon after, to satisfy demand, we saw the age of prints: various forms of printing did to painting what photography did to painting, and what posters did to both in the 20th. (FWIW, my favorite work is etching and mezzotints in particular, though my divorce lost me the small collection I’d had.)
    Early 1500s–1600s: Chiaroscuro woodcuts and multi-block color printing. After basic woodcuts, techniques like chiaroscuro (multiple blocks for tones and shadows, pioneered ~1516 by Ugo da Carpi, influenced by Titian/Raphael) allowed color and depth without full hand-painting. Made “painterly” effects reproducible at scale—further eroding demand for custom illuminated or hand-colored works. Early step toward color mass-production.
    1700s, the camera obscura in the seventeen hundreds let Rembrandt cheat on perspective. Guys who couldn’t draw at all started selling “realistic” scenes. No one called it cheating; they called it genius.
    Late 1700s–early 1800s: Lithography (invented 1796–1798 by Alois Senefelder). Drew directly on stone for fast, cheap runs—perfect for posters, caricatures, illustrated books, art prints. Killed slower intaglio/engraving for commercial work; made “fine art” reproductions accessible to bourgeoisie. Traditional engravers lost commercial gigs; many shifted to fine-art etching or teaching as lithography took over illustration and popular prints. Critics called it vulgar at first, but it exploded the market (Delacroix, GĂ©ricault used it).
    1840’s, calotype prints – same panic. Miniaturists who did cameos on ivory? Dead overnight. Nobody wants a teeny ivory profile when mom can sit still for two minutes and get a sepia tint. Ivory miniature painters and cameo carvers saw livelihoods vanish; many adapted to hand-tinting daguerreotypes or quit the field entirely.
    In the 1850’a photography had to evolve to aesthetics and the entire realism industry evaporated. (1874) Critics hated it. So did critics. But photography stole it’s job. Realist painters lost ground on hyper-accurate depiction; many pivoted to impressionism or abstraction as mechanical reproduction captured “truth” faster and cheaper. Portrait and landscape realists faced declining commissions.
    Mid-1800s: Stereoscopic photography and 3D views (1850s–1890s). Paired stereoview cards created illusion of depth from photos—mass-produced travel, educational, and novelty scenes. Landscape and architectural painters who sold detailed vedute (topographical views) to tourists lost market; many pivoted to looser styles or abandoned realism entirely as mechanical 3D “captured” scenes faster and cheaper.
    Mid-1800s: Industrialization & Arts and Crafts backlash. After photography stole realism, steam-powered presses and factories mass-produced decorative arts—wallpaper, furniture inlays, ceramics, textiles. Machines replaced hand-crafters in ornamentation. William Morris and Arts & Crafts (1860s–1910s) rebelled, insisting on handmade quality against “debased” machine goods. Direct response to the pattern: tech enabled cheap volume → quality declined → elite revival of craft (which then got commoditized anyway). Hand-weavers, woodcarvers, and ornamental workers lost factory jobs; movement tried to restore dignity through guilds but remained niche and expensive.
    1920s, we were in the same place after 1900 when economic center of the world moved to the USA, and more so after WW1 when art moved from mansions to apartments. Artists couldn’t make money at high investment art production. Hence the decline in representationalism and even impressionism. You had to turn out more volume in shorter time at lower cost.
    In the turn of the 20th with movies emerged to great fanfare (and a great sucking sound of artists into the industry over time). Opportunity knocks, shifting the employment of the whole industry.
    1920s–1930s: Art Deco and machine-age design. Streamlined chrome, Bakelite, and factory production replaced hand-crafted ornament in furniture, jewelry, graphics. Artisans in metals, inlays, and fine detailing faced obsolescence; many moved to industrial design or teaching, while the style celebrated mass production over individual craft.
    1920’s, color lithography. Suddenly Van Gogh posters hang above every couch in Paris. Originals rot in attics; the “work” drops to a buck. Traditional printmakers and painters saw mass posters undercut unique sales; many turned to fine-art limited editions or commercial illustration to survive.
    We went thru another change in the 50’s and 60s when art fell into the popular sphere and became a lower class interest that was easily commercially exploited.
    1950’s, silkscreen pops – Warhol prints soup cans on canvas. Critics screamed “sell-out,” but the kid in the Midwest got a Marilyn for twenty-five. Same product, zero labor.
    1965, offset lithography and four-color presses. Now the image quality rivals oils, and you can crank out twenty thousand units. Galleries panic, stock up on “originals.”
    1970’s to 80’s we saw the poster revolution that impacted the prior industry of prints. (My company sold tons of art slop… but at least at the time we were selling posters of previously quality work.)
    1980s-90s, desktop publishing. You used to be able to make a living in commercial art producing paintings, air brushing, even typesetting and past-up. My girlfriend in college had a job as a paste-up artist making those ads for neighborhood newspapers. My company made a fortune in the printing industry. But that’s all gone now. Air-brush artists die out. My girlfriend’s paste-ups become “vintage.” Comic inks get digitized, suddenly Manga’s printed in China at a nickel a page.
    1990’s, with emerging digital art – beforehand there were very few art books by comparison. same with things like comic books and illustrated novels. Exploded in volume.
    Once Chinese labor came onto the art movement in the 2000’s you could buy hand painted mantle-scale pieces for next to nothing. I bought a hunting scene with horses and dogs – but the people had slanted eyes. Made me laugh.
    Right now we are seeing digital artists panic because ai is replacing them in the low end, by enabling others to replace them at nearly zero cost in money and time. Yet others are thrilled at the opportunity to produce deeper creativity than just craftsmanship.
    Today, AI. Same joke – only this time the client never even sees a human wrist. The cheapest horse on canvas? Doesn’t matter if the eyes are slanted; the algorithm already fixed it.
    Broader recurring consequence: the pivot from craft mastery to concept/idea. Across eras—illuminators to printers, realists to impressionists, engravers to photographers—manual virtuosity loses value when tech handles replication. Survivors emphasize originality, narrative, spectacle, or curation (e.g., Duchamp’s readymades skipping craft; Warhol embracing multiples). The pattern: tech commoditizes skill → art becomes less about labor, more about idea or access → new elites emerge in concept over execution.
    IN OTHER FIELDS
    The same thing happens with programmers. You play the learning curve game long enough to profit from the technologies you invest in, then you stop paying the learning curve game because you can’t do it any longer, and so milk the de-adoption curve while better guys move on, and worse guys leave the industry.

    Nothing New In The AI Age

    The pattern we’ve traced—from illuminated manuscripts crumbling under the printing press, to realist painters sidelined by photography, to airbrush and paste-up artists vanishing into desktop publishing, to today’s digital illustrators watching AI commoditize low-end work—isn’t a modern crisis. It’s the eternal rhythm of art’s economy. There’s nothing new here under the sun.

    Every leap in reproduction technology has done the same thing: it floods the market with cheaper, faster, more accessible versions of what used to demand rare skill, time, and patronage. The “aura” of the unique original—whether a hand-illuminated page, a one-off portrait, or a labor-intensive digital piece—gets diluted when copies (or near-infinite generations) become the norm. Walter Benjamin nailed this in 1935 with mechanical reproduction: art sheds its ritualistic dependence on the singular, the cult object, and becomes designed for mass dissemination. The work isn’t destroyed; it’s democratized, commoditized, and often decoupled from the artist’s hand. Scribes became proofreaders or faded away; miniaturists tinted photos or quit; realist portraitists pivoted to looser, more expressive styles; commercial illustrators migrated to concept, curation, or new niches like teaching the tools they once wielded.
    The human cost repeats too: livelihoods shrink for those tied to the old high-investment craft, panic spreads (“this will kill jobs for artists”), critics decry the loss of soul or authenticity, and elites mourn the debasement of quality. Yet the field doesn’t die—it expands in unexpected directions. New roles emerge (printmakers, photographers-as-artists, prompt engineers), new audiences grow (merchants buying Madonnas, bourgeoisie hanging litho posters, kids remixing AI outputs), and creativity finds fresh ground by emphasizing what machines can’t replicate: idea over execution, narrative over fidelity, spectacle over sweat, or sheer originality in concept. Impressionism bloomed when photography stole hyper-realism; Pop art embraced multiples when lithography made uniqueness quaint; conceptual art skipped craft altogether with readymades. Survivors adapt by leaning into the very disruption—turning the tool into the medium, the copy into commentary, the cheap abundance into deeper expression.
    AI is just the latest iteration: near-zero marginal cost for visuals that once required years of training. Low-end commercial work gets automated first (stock images, simple ads, mantle-scale decor), just as engraving undercut fresco copies, lithography killed hand-colored illustrations, and Chinese factories buried mid-tier hand-painting. The panic feels existential because it always does in the moment. But history shows the arc: tech commoditizes skill → supply explodes → prices crash for rote labor → artists pivot to what remains human (emotion, critique, novelty, curation, performance) → the economy reshapes around new forms of value.
    Nothing is lost forever; the discipline evolves. The horse on canvas with slanted eyes still sells if the story or the vibe lands. The algorithm fixes the eyes, but it can’t invent the why. Art persists because humans do—adapting, complaining, innovating, and outlasting every tool that promised to replace them. The disruption isn’t the end of art’s economy; it’s the engine that has driven it forward for centuries. Same cycle, shinier gears.

    There is nothing new about this turnover in the art industry. It’s as ruthless as any other craft that is subject to the possibility of technological innovation that expands the market by the reduction of prices.

    Cheers
    Curt Doolittle
    The Natural Law Institute
    And Runcible Inc.

    Source date (UTC): 2026-02-04 03:41:22 UTC

    Original post: https://x.com/i/articles/2018892612527354072

  • That’s straight thinking. But I’ll give you a suggestion: there is no moat. They

    That’s straight thinking. But I’ll give you a suggestion: there is no moat. They have no competitive advantage. AI will be as commoditized as email. We will all pay for hardware devices once prices come down further and all have our own AI’s probably on our phones, and there will be no money to be made by these large companies so much as they will serve large corporate and government clients like amazon and microsoft do today.
    Prices are collapsing, data center expansion is being reduced. That’s because they already know.
    The problem is they can’t afford to not play the game.

    Source date (UTC): 2026-02-03 19:04:22 UTC

    Original post: https://twitter.com/i/web/status/2018762503661203596

  • QUESTIONS: 1) Has there been any significant employment impact due to AI or is i

    QUESTIONS:
    1) Has there been any significant employment impact due to AI or is it that companies are downsizing and using ai as the justification. i ask because there is little evidence of ai adoption affecting businesses so far.

    2) What jobs were replaced by the sequence of computer revolutions, and what jobs were created?

    ANSWERS

    Source date (UTC): 2026-02-03 16:29:41 UTC

    Original post: https://twitter.com/i/web/status/2018723576166429160