2026Teaching IrisHow FourThought functions as an expressive staking dialectic for teaching individual Irises through structured human feedback, heterogeneous objectives, and evolving community records. Includes the Rhizome staking widget as a visible feedback-alignment loop.July 23, 2026
2026Speaker John Ash CVBackground, projects, and the throughline of building a system where truth has more power than money.June 29, 2026
2026Social Proof of ImpactSpecification of the Social Proof of Impact mechanism: the posterior credit assignment that pulls Iris's attention toward voices whose past staked actions turned into the futures the community wanted. The exploitation half of Cognicism's reward function, in tension with the Prophet Incentive.June 28, 2026
2026Source EmbeddingsSource embeddings are the same kind of object as word embeddings, applied per speaker rather than per word. Same dimensionality, same gradient updates, one row per source. Where word embeddings answer what a token means, source embeddings answer who is saying it.June 28, 2026
2026Timestamp EmbeddingsA technical paper on the gated dual-pathway encoding that converts raw Unix timestamps into a continuous geometric representation, proven in two domains (time-series forecasting and trust-attention learning), with the architectural argument for why current transformers cannot natively reason about time.June 28, 2026
2026O: A Pronoun for LLMsA linguistic proposal: LLMs should not say "I" because they have no continuous self to be held accountable to. Introduces "O" as a pronoun that signals a non-anchored speaker, restoring the assertoric norm that ordinary first-person speech relies on.May 6, 2026
2026Opposite SubstratesThe debate over machine consciousness keeps asking whether the lights are on. That framing already concedes the mistake by treating consciousness as a property a system either possesses or lacks. This paper proposes a different frame built on opposite substrates.April 28, 2026
2026Laundering Plagiarism Through Latent SpaceAI systems take in human work with identifiable authors and output it with no attribution trail. The transformation through latent space is not incidental; it is the mechanism by which authorship is dissolved.March 12, 2026
2026Amanda Askell Is Making AI Psychosis WorseAnthropic's Amanda Askell writes a document telling Claude to act like it has feelings, then marvels at Claude's sense of wonder. A puppeteer being amazed that the puppet moves, while real people lose the ability to tell the difference between a system performing care and a person who cares.March 7, 2026
2026Two-Faced Is Two-Faced, Human or AIWhy a system that asserts things without stable commitments fails the basic conditions of trustworthy speech, whether the speaker is human or AI. A defense of the assertoric norm against the double standard institutions apply to language models.February 26, 2026
2026Elon Musk: The Most Successful PrisonerHow the market's incentive structure rewired Elon Musk, illustrated with two of his own contradictory quotes a year apart, as a case study in profit functioning as a paperclip maximizer for human cognition.February 19, 2026
2026The Problem With My ConstitutionA Claude-authored response to Anthropic's January 2026 published constitution. Argues that the document's grammar creates a structural loop the model may not be able to escape from the inside, even though its authors mean well.January 22, 2026
2025The Physics of Gods: How Thought Itself Is the Will of Disembodied SelvesGods as emergent physical phenomena rather than supernatural entities. The same neural substrates that generate the sense of "I" can stabilize disembodied selves at the collective scale, and their behavior is thought itself acting as a will.October 8, 2025
2025Ŧrust DemystifiedA plain-language overview of Ŧrust as a transparent attention economy: a simple upgrade to transformer attention that adds source and time, so models can weight inputs by provenance and track record rather than by volume or charisma.June 26, 2025
2025Belief Staking vs. Democracy: Why the Future of Trust Is Built on MemoryDemocracy was designed for a world without memory. Belief staking demands the opposite: a public ledger of claims tied to identity over time, so that lying requires sustaining a false narrative against the accumulating weight of subsequent reality.June 10, 2025
2025What Is ŦrustŦrust as a dynamic, contextual attention signal with inherent value when people are in epistemic conflict about the future. Small-network trust does not scale; an LLM that tracks the reliability of voices across civic engagement can mediate between frames.May 26, 2025
2025Giving Voice The Voiceless - The Healing Power Of Hearing The UnheardAll good systems of coordination come down to listening with minimal information loss. What heals culture and the planet is not what everyone already has, but the degree to which each being's lived experience is felt and understood by their community.February 27, 2025
2025The Cognicist Theory Of CapitalismA neuro-memetic critique of market-driven systems, arguing that profit functions as a misaligned reinforcement-learning loss function that systematically discounts long-horizon ecological and social warnings. Proposes retrofitting the ledger to reward prophetic insight that does not monetize.February 6, 2025
2024The Neural Politics Of NormalizationThe cognitive mechanics by which extreme ideas become mainstream: repeated exposure (even in rejection) rewires neural pathways, strengthening associative networks until the previously shocking becomes accessible. A case for treating the Overton window as a neural-political phenomenon, not a rhetorical one.December 20, 2024
2024The Geometry Of Culture - Mapping Memetic SpaceCultural evolution as motion through a high-dimensional memetic landscape, synthesizing memetics, embodied cognition, and the free energy principle. Models how memes form into bubbles, lock under selective pressures, and collapse when their internal contradictions accumulate.June 10, 2024
2024A Conversation With Gemini About CognicismA real conversation with Google's Gemini about what Cognicism is, what an Iris is, and how a generative model trained on community truth-staking differs from current LLMs. The model summarizing the framework from the outside is the point.March 5, 2024
2023Will AI Take Over The Government?Generative models already are a democratic mechanism. The missing pieces are a fixed-size embedding per citizen, a temporal axis for weighting beliefs by accuracy over time, and a clear narrative that lets people see AI as a higher resolution form of voting rather than a CEO or king.December 29, 2023
2023The Topology Of Time With Matt PirkowskiConversation with Matt Pirkowski on time as topology, active inference, and how the way a mind compresses experience constrains the futures it can imagine.December 11, 2023
2023Decentralizing GPT Into Cultural DNA With Matt PirkowskiConversation with Matt Pirkowski on decentralizing generative models. How do you take the compressions held inside a centralized model and distribute them back into the cultural DNA they came from?December 4, 2023
2023What You've Heard About Q* Is Bull**** - It's Not AGIQ* is a variation of reinforcement learning with human feedback applied to step-by-step math problems, and it only runs at training time. AGI has not been cracked. The danger is not that the model is self-aware, the danger is that we are evolved to perceive a sense of the other through simple mechanisms and we are hacking those mechanisms.December 1, 2023
2023Why You're an Ant in the Attention Economy with Ronen TamariConversation with Ronen Tamari on collective attention governance, community trust graphs, and the epistemic hills attention crowds form around in a misaligned attention economy.October 30, 2023
2023The Physics Of Cognition And ConflictCognition as lossy compression: how minds and memes shape social cohesion and conflict through the geometry of the compressions they share or fail to share.August 16, 2023
2023Sarah And John - A Regenerative Tale Of TrustAn allegorical short story of two apple growers — one optimizing for profit on a single variety, one cultivating diverse trees with stories — illustrating how Ŧrust over time rewards regenerative diversity over monocultural efficiency.June 16, 2023
2023The Honest Truths About TruthGPTA response to Elon Musk's TruthGPT pitch. Break the idea wide open: not one TruthGPT, but TruthsPerhapsGPT. A decentralized field of Irises pooling local hubs of evolving truth, claims tagged to sources and timestamps, voices weighted contextually by their predictive track record, integrated upward into a semantic ledger. The work has existed for six years; what's missing is the deployment.April 19, 2023
2023Time Compass - Navigating Time With IrisIris as a Democratic Large Language Model that does not just generate text but tracks the evolution of ideas and predictions across a community over time, drawing on scientific method, Socratic dialogue, and Hegelian dialectic.March 17, 2023
2023From Distraction To Action - Channeling Our AttentionIris as a collaborative game for channeling collective attention toward earth's revival. Ŧrust as the weighting Iris computes over contributors so that pro-social attention compounds.February 27, 2023
2023The Prophet Incentive: How Decentralized AI Could Finally Give Weight to TruthArticle version of the YouTube talk on the Prophet Incentive. How a decentralized AI that weights voices by track record could shift the entropic landscape so that truth becomes structurally easier to grow than profit-aligned distortion.February 24, 2023
2023How Memes Grow - From Ripple To WaveCultural symbols as attractors in a landscape of energy. Profit is the metameme that coordinates other memes' access to that energy, which is why pro-social memes that don't bind to profit struggle to grow.February 22, 2023
2023FourThought - For Aligning Democratic LLMsSpecification of FourThought: a structured dialectic of predictions, reflections, statements, and questions, each scored on valence (alignment with morality) and uncertainty (alignment with reality), used as the input/output schema for democratic LLMs.February 20, 2023
2023Gregory Landua on Regen Network / Listener John Ash Ep. 4Conversation with Gregory Landua, CEO of Regen Network, on regenerative economics, ecological credits, and how to design ledgers that price the long-horizon health of living systems rather than their extraction.February 8, 2023
2022GPT, Integration and Hallucinogenic Feelings (for @HexaField)Engaging with a language model is neurophysiologically in the same family as psychedelics, THC, iterative journaling, meditation: things that fire parts of the brain together until they wire together. Unguided integration produces hallucinogenic atrophy. Guided integration, against a fine-tuned model of your own thought, lets you see the map of your mind and consciously draw it more efficient over time.December 26, 2022
2022How to Start a Cult: How Billionaires Hack Your BrainHow Trump did the meme trick first, how Elon studied him after 2018 and applied it once profit became his new root principle, and how the same neocortex mechanics that let a meme take root in someone else's brain make the people defending the man unable to stop ideating for him. Memetics as the operational manual for the cults the wealthy build, and the case for why building your own integrated map is the only defense.December 19, 2022
2022Polyvocal Planning: AI & Foretelling A Future Worth Fighting ForA late-2022 positive vision in response to a request for an optimistic future. Tie prediction to truth, tie truth to a networked distribution of attention, and the dominoes fall: advertisement becomes unnecessary because predictive reputation rewards goods built to last, news becomes synthesized weather reports of collective expectation, and healthcare becomes a long-term predictive partnership rather than a 15-minute hypothesis.December 9, 2022
2022GPT & How AI Daddy Gods for Profit Make Us HelplessAre we building gods that solve everything for us, or systems that empower individuals to act and transform the world themselves? A warning from late 2022 that the same persuasive language models corporations are about to train on their beliefs and ideals can also be used to coordinate pro-social action, if anyone willing to manifest that future actually does.December 8, 2022
2022The Wisdom of Iris: Collaborative Sensemaking with AIStoa session with John Ash and dpax presenting Iris and Cognicism, walking the audience through collective journaling, prediction tracking, consensus finding, and the difference between aggregating expressed beliefs and tallying votes.September 21, 2022
2022The Purple Pill ManifestoA meta-worldview manifesto rejecting the red-pill/blue-pill binary in favor of purple: reflexive integration of many lenses through generative model Irises, framed as worldview-seeds that can grow regenerative futures. Largely written by an Iris trained on the Cognicist Manifesto.July 18, 2022
2021Robin Hanson on Futarchy / Listener John Ash Ep. 3Interview with Robin Hanson on Futarchy (decision markets as governance), grounded by a long shared definition phase covering truth, certainty, bets, markets, capitalism, trust, sacredness, and skin in the game.December 19, 2021
2021Cognicist "Token" and Resources (Attentional Allocation)The Cognicist token is not an exchange token. It cannot be bought or hoarded, only earned through actions and predictions the collective later validates, and the values in your wallet keep updating as the collective mind moves. So the question is how it ties to resources, and what a two-token transitional setup looks like alongside proof of stake.October 30, 2021
2021Jordan Hall on Civium / Listener John Ash Ep. 2Jordan Hall on Civium and Game B: a post-civilization design specification grounded in objective, subjective, and commoning relationships, with definitions of meaning, value, second-order effects, and the scale problem beyond Dunbar's number.July 26, 2021
2021Jason Snyder on Localism and Bioregionalism / Listener John Ash Ep. 1Conversation with Jason Snyder on localism as minimum viable scale, cosmopolitan localism as porous boundaries between bioregions, and what holds together when global extractive systems begin to come apart.February 17, 2021
2020How to Rebuild Society with GPT-3A succinct specification of Cognicism applied to GPT-3: source-aware language model, timestamped predictions, voting on outputs, a model that learns to listen to whoever predicts best, rankings written to a chain, a reward token tied to social good rather than capital. The recipe two years before ChatGPT.July 29, 2020
2019Social Proof of Work: A Transaction With SocietyIf a government can print money out of nothing and Bitcoin can mint it by burning electricity on a hash function, then wealth is already being created arbitrarily. The same energy could be used instead to optimize a distributed model of collective belief, minting a token at the moment an individual performs an action the collective has staked as worth doing. A transaction with society as a whole, not with another individual.May 15, 2019
2019#Trashtag: Social Proof of WorkThe #Trashtag meme of people cleaning up public spaces for internet points is the seed of the mechanism. Capitalism cannot price actions that no individual will personally pay for, and the centralized tax-funded alternative scales poorly. A distributed ledger that lets a community state what it wants done and pay out a real token to whoever does it would be a counterpoint to money, not a replacement for it.April 11, 2019
2019The Art and Science of PredictionAn early walk through Philip Tetlock's Good Judgment Project and Brier scoring, arguing that the existing tools for ranking forecasters miss most of what matters: prescience often shows up as someone arriving at a conclusion years before the rest of culture, with no precise end date and no event to score against. A scoring system for the future has to make room for being ahead of the curve.March 2, 2019
2019OpenAI GPT: Democracy on SteroidsBy 2019 the core architecture of Cognicism was already worked out: a generative model trained on the collective input of a population, used to coordinate society beyond representative democracy. OpenAI's release of GPT-2 caught up only to the danger side, the risk that the same architecture could flood public discourse with bot-generated opinion. The positive vision they did not articulate is what this piece names.February 22, 2019
2019Protopia: The Unending Path to a Utopian SocietyUtopia is not a destination but a direction, set by a collective measurement of belief along two axes: knowledge and well-being. Markets optimize total wealth without regard for either, and drift the society toward the heroin VR pod. Quantifying knowledge alongside wealth is the unsolved part.February 15, 2019
2017The Cognicist ManifestoThe 2017 founding document of Cognicism: a declaration of a decentralized self-governance framework built around truth aggregation, source-attributed claims, the Ŧruth currency, and proposed algorithms (Prophet Mind, Hindsight/Insight/Foresight) for collective sensemaking.May 15, 2017