{"path":"vision.md","content":"# Deliberus Vision\n\n**Weaving our minds together.**\n\n## Core Intent\n\nIndividual minds, no matter how bright, are not up to the task of engaging with the full complexity of any wicked, polarizing, or multifaceted issue. As humans, we tend to outsource our thinking to gurus, strongmen, public intellectuals — to worship the individual intellect. The key to advancing as a species is engaging more in productively weaving our intellects together, and extracting into software the criteria for what constitutes well-founded knowledge and wisdom.\n\nDeliberus is infrastructure for this. Not a forum, not a debate tool — a civilizational reasoning substrate where any human can navigate any topic through any worldview's lens, at any level of complexity, to understand not just what people believe but *why*. The analytical tools decompose arguments into their logical bones; the [attunement layer](analysis-and-attunement.md) lets you put on different worldview glasses and see what you've been missing. Neither alone is Deliberus — the analysis serves the attunement, and the attunement motivates the analysis. See [What This Could Enable](civilizational-vision.md) for the full civilizational implications.\n\n### The Problem\n\n- Bias, propaganda, perverse incentives, polarization, echo chambers, clickbait sensationalism, \"algorithm-fueled limbic hijack\"\n- Information overload without tools to tame it\n- Personal strategies for rationality are costly and fragile — collaborative tools scale better\n- Social media optimizes for engagement, not truth-seeking\n- Democratic institutions lack technological scaffolding for collective sense-making\n- AI safety requires better human decision-making infrastructure\n\n### The Solution\n\n- Make arguments, counterarguments, premises, conclusions, definitions, evidence, values, and \"ought/should\" claims **explicit and navigable**\n- Surface hidden social dynamics: tone/emotion analysis to detect self-censorship, obfuscation, or social forces distorting truth-seeking\n- Add \"scoreboard\" elements — like \"Wikipedia + experts putting claims in a scoreboard\" (Joshua Vantard exchange)\n- Leverage LLMs for scaffolding + humans for grounding (modern synthesis)\n- Build rationality standards into the software itself, not relying on users to self-enforce\n\n### The Aspiration\n\n> \"The world would benefit immensely from a web-based platform to allow an unlimited amount of participants to examine visually and contribute to debate in a non-linear fashion, minimizing complexity, noise and duplication, and enhancing the results of collaboration.\"\n> — Deliberus description (2012, partly inspired by conversations with Dustin Juliano)\n\n> \"We need new ways to use our collective intelligence to tame information overload. Finding truth is hard. Personal strategies for rationality are more costly and fragile than strategies that we can build into tools that we use collaboratively.\"\n> — Deliberus P2P Foundation wiki page (2013)\n\n> \"A system of (argument map + human society) might be a hybrid AGI.\"\n> — Ivan Phillips (Dec 2011), in conversation with Fredrik\n\n> \"I'd like to find some algorithm for humanity to reason and make decisions together in a way that combines the democratic elements and general efficiency of a decentralized system like markets with the discerning force of logic, the common ground of science, technology, specifically the internet and software to amplify the powers of human language.\"\n> — Fredrik (2012, Simplenote article draft). See [research/simplenote-archive-analysis.md §5](research/simplenote-archive-analysis.md)\n\n> \"I see it as a spreadsheet for knowledge rather than an expert system.\"\n> — Tom Boniecki (Jan 2011, responding to Ivan Phillips's \"A Tool to Industrialize Knowledge\" note). See [research/simplenote-archive-analysis.md §3](research/simplenote-archive-analysis.md)\n\n## A New Medium, Not Just a Better Tool\n\nThe 2012 Leverage Research notes contain a claim stronger than anything else in this document: improving discourse granularity from coarse to fine — from paragraph-level debate to premise-level structured argumentation — **\"could be likened to the advent of language itself.\"** Language enabled fine-grained communication between minds; Deliberus enables fine-grained *reasoning* between minds. This reframes the project from \"a better debate tool\" to a new medium for collective cognition. (See [research/simplenote-archive-analysis.md §1](research/simplenote-archive-analysis.md))\n\n## The Evolutionary Frame\n\nEvolutionary biology suggests the next big step is collaboration at a higher level of organization. This pattern repeats: the birth of multicellular life, the emergence of culture, the formation of civilization. Deliberus is the next instance — a platform that lets humanity's collective reasoning exceed any individual's capacity, the way the scientific method lets collective inquiry exceed any single scientist's insight. \"We are shifting to a collaborative default as a species, but it looks like we need a better framework for it.\" (See [research/simplenote-archive-analysis.md §5](research/simplenote-archive-analysis.md))\n\n## Non-Zero-Sum Discovery (Apr 6, 2026)\n\nThe newer cooperation research sharpens the evolutionary frame: Deliberus is not just a map of disagreement, but infrastructure for discovering hidden non-zero-sum games. Wright's *Nonzero*, Deutsch's error-correcting epistemology, Coasean transaction-cost economics, Ostrom's commons governance, deliberative democracy evidence, false-polarization research, and cooperative AI all converge on one operational claim: much apparent zero-sum conflict is positive-sum possibility hidden by semantic confusion, information asymmetry, narrative warfare, and high costs of mutual understanding.\n\nThe implication for product strategy is concrete. Contested concept detection, decomposition, bridging signals, worldview lenses, and the persistent graph all lower the transaction cost of understanding why another person or group holds a position. That is not merely \"better discourse\"; it changes the game being played by making cooperative options visible. This research converges on a conviction that predates it here and is stated in its own voice below: [Opacity Is a Cost, Not a Mystery](#opacity-is-a-cost-not-a-mystery-aug-2026). Full synthesis: [research/non-zero-sum-economics-and-civilizational-cooperation.md](research/non-zero-sum-economics-and-civilizational-cooperation.md), with companion research on [game theory and cooperation economics](research/non-zero-sum-game-theory-and-cooperation-economics.md), [trust and false polarization](research/trust-economics-and-false-polarization.md), and [AI-augmented cooperation infrastructure](research/ai-augmented-cooperation-infrastructure.md).\n\n## Philosophical Constraints (Anti-Goals)\n\n- **Reduce taboo pressure / social masking** that prevents direct truth-tracking — not culture-war framing, but stopping the punishment of honesty\n- **Reduce duplication and repetition** — surface adjacent relevant material so people don't re-litigate mapped territory\n- **Truth-seeking over persuasion** — unlike traditional debate focused on winning\n- **Post-polarization**: tools to hold dialectical tension and synthesize opposites, \"both/and synthesis\"\n- **Hyperhumanist**: tech that extends human essence rather than replacing it\n- **Not naive about ethics**: \"felicific calculus?\" with caution — \"more complicated than that\"\n\n## The Deeper Thesis (Mar 27, 2026)\n\n**Individual minds, no matter how bright, are not up to the task of engaging with the full complexity of any wicked, polarizing, or multifaceted issue.** As humans, we have a tendency to seek clarity and guidance from gurus, strongmen, public intellectuals, shamans, grandfathers — to worship the individual intellect. The key to advancing as a species is engaging more in productively weaving our intellects together and, as far as possible, extracting into software and agreed-upon algorithms the criteria for what constitutes well-founded knowledge and wisdom, including well-evidenced opinions.\n\nThis is not about replacing human judgment with computation. It's about building infrastructure that lets collective reasoning exceed any individual's capacity — the way the scientific method lets collective inquiry exceed any single scientist's insight, but generalized beyond empirical science to ALL domains of human deliberation.\n\n**Progressive disclosure of complexity**: Adoption hinges on the user choosing their depth. A casual user votes, reads summaries, trusts the system's epistemic scaffolding. A curious user expands argument bundles, examines premise chains, checks evidence quality. An expert drills into probabilistic weights, falsification history, causal counterfactuals. The same underlying graph, rendered at different cognitive resolutions. Trust can be earned here like nowhere else — because the reasoning is transparent, auditable, and structured all the way down.\n\n## The Core Dialectic: Analysis ↔ Attunement\n\nTwo drives inhabit this project: *ana-lysis* (Greek: \"loosening apart\") and *com-prehendere* (Latin: \"grasping together\"). The analytical pole decomposes, verifies, computes. The attunement pole inhabits other worldviews, understands WHY, grasps the whole. Pure analysis without attunement → the \"neutrality illusion.\" Pure attunement without analysis → consensus without rigor. The synthesis — attunement-informed analysis — is the soul of Deliberus. The dialectic runs through every layer: philosophy, strategy, architecture, UX, data model.\n\nAt civilizational scale, the attunement pole is what makes [love through understanding](civilizational-vision.md) possible — not agreement, but the recognition that the person across from you is reasoning from different premises, not from irrationality.\n\n**Full exploration**: [analysis-and-attunement.md](analysis-and-attunement.md) — the two poles, the synthesis, McGilchrist's hemispheric thesis, Žižek's challenges, and why the dialectic IS the architecture. The [worldview filter](worldview-lenses.md) has its own dedicated doc.\n\n## Opacity Is a Cost, Not a Mystery (Aug 2026)\n\nBefore the wager, a conviction. The difference between those two words is the whole point: the wager is what I am betting, and this is what I am sure of.\n\nWhen someone holds a position I find baffling, the reason they hold it is *findable*. Not always agreeable, not always sympathetic, but findable. The distance between two minds is traversable. What stands in the way is not a mystery about other consciousnesses or an unbridgeable difference in kind. It is a **cost**: the work of finding out what someone actually means, which premises they are reasoning from, and which of those they would still defend under pressure. Costs behave differently from mysteries. A mystery you contemplate. A cost you lower, and you lower it by building something.\n\nEverything else here is an instance of that. [Depth](depth.md) lowers the cost by decomposing a claim until the place two people part ways becomes visible. [Lenses](worldview-lenses.md) lower it by letting a reader stand inside a frame that is not theirs. [Attunement](analysis-and-attunement.md) is the half of the cost that no argument can pay. [Bridging](bridging.md) is a signal that only becomes detectable once the cost is low enough to measure. [Convergence](convergence.md) is the wager about what is found at the bottom, once the descent is possible at all — which is why this section comes first. Convergence cannot be tested unless the descent is traversable. The claim below is my boldest; this one is my strongest.\n\nTwo things it does not say. It does not price away attunement: understanding another person is partly analysis and partly an encounter, and if \"cost\" ever became the master frame here, the analytical pole would have eaten the other one, which is precisely the failure [the core dialectic](analysis-and-attunement.md) exists to prevent. And it does not claim that all disagreement is confusion. Lowering the cost *dissolves* disagreement that was confusion, and only *clarifies* disagreement that is conflicting interests. Someone who understands me perfectly and wants what I have was never a comprehension problem. Clarity is still worth a great deal in that case, because you stop spending years on a disagreement that was never the real one, but it is not resolution, and a tool that promised otherwise would be lying about the world.\n\nEconomics has a name for the structure of this claim: Coase showed that transaction costs decide which cooperation is possible at all, and the same constraint turns out to have been named independently at nearly every scale, from a single bundled premise to relations between states ([the scale ladder](research/fractal-scales-and-temporal-frame.md), [the cooperation synthesis](research/non-zero-sum-economics-and-civilizational-cooperation.md)). I did not get the conviction from the literature. The literature is how I know it is not only mine. And the name is quietly doing its own work: \"opacity\" is a slightly formal word, so some readers will pause on it, and that pause — the half-second of looking it up, or inferring it from *cost, not mystery* — is the thing itself in miniature. A small opacity, a small cost, willingly paid, and then it is gone.\n\nOne consequence of the conviction is now stated as a precondition rather than a hope (ratified 2026-08-18): the promise is **not faster decisions — it is checking that keeps pace with persuasion**. Machine persuasion wins on sheer throughput, so speed is the adversary's advantage too; what this infrastructure adds is verification that scales at the same rate as claim-production, arriving later and per claim ([the scrutiny gap](research/the-scrutiny-gap.md)).\n\n## The Founding Wager (July 2026)\n\nI believe the vast majority of people have good hearts. They interpret what's feasible and wise differently — through different experiences, different information, different meanings hiding in the same words — but the care underneath is broadly shared. That belief is where Deliberus comes from, and it sharpens into a wager I hold fallibly: **decomposed deeply enough, honestly enough, our views converge.**\n\nNot because nothing real divides us. Ask *\"what exactly is being weighed against what?\"* of any value conflict and the vague wall of \"we just weight things differently\" opens up: into beliefs about consequences, postures toward risk, choices about whose interests count, rules for adding things up. Each part can be argued on its own. Most of the disagreement dissolves on the way down; what remains at the bottom, when anything remains at all, is a [typed residue](research/convergence-wager-typed-residues-synthesis.md) — small, precisely stateable, and still arguable — no longer fog, but an object two people can point at together.\n\nTwo convictions ride along. Better-informed and better-reasoned are real distinctions: views can be ranked by how well they survive scrutiny, validated against reality wherever reality answers, with the ranking rules themselves standing in the graph as challengeable claims. And whichever way the wager goes, the work pays, because legibility is the guaranteed floor: every disagreement that passes through the system comes out precise, typed, and shared. \"We share more than divides us\" and \"we share less than we hoped\" are both discoveries, and the graph must be able to report either one. I want to win this wager — and the instruments are built to tell me if I'm losing it.\n\nThis is also why Deliberus is not, finally, a product, a personal tool, or a research instrument. Those are phases it passes through. It is **public epistemic infrastructure**: a commons in the lineage of Wikipedia and open proof libraries, carrying a falsifiable civilizational wager.\n\n**Which half the analogy carries (Aug 18, 2026):** Tested against how complexity reduces at other scales, the wager splits into two claims that turn out to have very different support. *The residues are few and narrow* is close to a universal pattern — endless substances reduce to about a hundred elements, all languages to small phoneme inventories, physics keeps finding fewer things underneath. *There is a shared bottom* is the weaker half: chemistry bottoms out in a hundred elements rather than one, and all life shares one genetic code **because of common descent**, not because the code was forced. So what recurs across scales is convergence onto *a small basis*, not onto *the same basis* — and the headline needs the second. The shared-bottom half does have a real argument, but it is phylogenetic rather than decompositional: shared evolutionary psychology, moral foundations recurring cross-culturally, Curry's seven cooperative moral rules reported across sixty societies. That is an inherited-substrate case, and it is not the case the corpus currently makes. Worth knowing which half is standing on what, since the rider is better supported than the headline. Full test: [research/convictions-across-scales.md](research/convictions-across-scales.md) § 2.2.\n\n## Semantic Disambiguation as Root Cause (Mar 28, 2026)\n\nSome words mean different things to different people — and **this is one of the primary sources of disagreement in the world.** \"Freedom\" to a libertarian, a socialist, and a Buddhist are three different concepts wearing the same word. The Swedish \"barnfattigdom\" debate was entirely semantic — everyone agreed on facts, but different operative definitions generated a public fight about nothing. Deliberus addresses this structurally: definitions are first-class argument nodes, not metadata. The moment a user sees \"oh, we were using the same word differently — THAT'S why we disagree\" is one of the most powerful the platform offers. At scale, this is how [tribal epistemics dissolve](civilizational-vision.md) — most disagreement lives in the middle layers (definitions, cultural framing), not in genuine value divergence. (See [research/semantic-disambiguation-and-concept-tracking.md](research/semantic-disambiguation-and-concept-tracking.md), 2013 concept tracking sketch in [sketches.md §Sketch 7](sketches.md))\n\n## One Graph, Not Separate Canvases\n\nThere is no \"personal mode\" and \"social mode.\" There is ONE shared deliberation graph. You approach it from your perspective, contribute to it, and can shift perspectives to see it differently. Like Wikipedia: you come to read, gradually start editing. Creation and exploration are the same act. \"Anyone can go to Wikipedia and create a new page — why would individuals need an empty canvas, if we all want to follow the light of reason?\" This single shared graph is what makes the [evolutionary next step](civilizational-vision.md) possible — improving discourse granularity from coarse to fine, from paragraph-level debate to premise-level structured argumentation, \"could be likened to the advent of language itself.\"\n\n**Conversational read/write loop (Apr 8, 2026):** The same one-graph principle applies when a user enters only a question, topic, or short half-formed thought. Deliberus should read from the shared graph to answer at the user's level, but should not silently publish the user's raw prompt as graph ontology. The graph-deepening move is provisional first: capture the utterance, infer possible claims/questions/terms/evidence needs, and ask for lightweight confirmation before durable graph writes. See [research/truth-graph-evidence-system.md](research/truth-graph-evidence-system.md).\n\n**Beginner-readable first take (Apr 8, 2026, later):** That conversational loop only works if the first screen feels like help rather than system diagnostics. The opening response should use ordinary language a newcomer can parse immediately: a first answer, a few related earlier ideas, the main uncertainty, and one clear next move. The inspectable graph remains underneath, but the product should not demand that the user learn Deliberus's internal vocabulary before benefiting from it. See [research/truth-graph-evidence-system.md](research/truth-graph-evidence-system.md), [ux-principles.md](ux-principles.md).\n\n**One graph, staged maturity (Apr 8, 2026, later):** The right long-term answer is not a clean split between \"conversation outside the graph\" and \"proper graph content inside it.\" Deliberus should preserve rough deliberation early and refine it in place over time, but with explicit maturity and provenance: raw trace, draft structure, candidate structure, canonical structure, and superseded structure should not all count the same. The graph stays one graph; what changes is how much epistemic weight different parts of it carry by default. See [research/truth-graph-evidence-system.md](research/truth-graph-evidence-system.md), [research/soft-canonical-clustering-and-reversible-merge-semantics.md](research/soft-canonical-clustering-and-reversible-merge-semantics.md).\n\n**Share the code, compartmentalise the contents (Aug 18, 2026):** Tested against other scales, this conviction comes back neutral in a useful way. For a single shared substrate: one genetic code, one price system, one internet address space, one Wikipedia. Against it, and strongly: **evolution invests enormously in boundaries.** The eukaryotic cell is *defined* by internal compartments — membranes, organelles, the blood-brain barrier, tissue boundaries, speciation itself. Life does not pool. It compartmentalises aggressively while sharing a code. What is universally shared is the *code*; what is universally not shared is the *contents*. So staged maturity is not a compromise of the one-graph principle — it is the form that principle takes at every level where it actually works, and the membrane is load-bearing rather than a concession. Full test: [research/convictions-across-scales.md](research/convictions-across-scales.md) § 2.8.\n\n**Implementation note (Apr 8, 2026, later evening):** The first behavioral slice of this is now live: default public retrieval can distinguish staged maturity instead of using recency as a crude proxy. Rough saved material does not have to vanish, but it also does not automatically become the thing a newcomer sees first. The next persistence slice is live too: short-authored first takes can now be preserved as non-public deliberation drafts containing the raw exchange and provisional promotion hints. The harder next question is how promotion should actually happen across turns and over time. See [research/truth-graph-evidence-system.md](research/truth-graph-evidence-system.md).\n\n## Depth: No Copout Axioms, No Permanent Bottom\n\nMost things people call \"axioms\" are conclusions that haven't been examined deeply enough. The graph has no permanent bottom — only a current bottom that's always invitable to go deeper. The convergence thesis: when you DO keep decomposing, most disagreement dissolves — and what survives is not a wall of \"different values\" but a small set of typed residues: precisely stateable, still arguable, and countable (see The Founding Wager above, and [research/convergence-wager-typed-residues-synthesis.md](research/convergence-wager-typed-residues-synthesis.md)). The wager, held fallibly, is that the residues are few, narrow, and shrinking. Reason, like Dennett's Universal Acid, dissolves assumptions regardless of topic. Tao's verification thesis validates this architecture from mathematics.\n\n**The acid dissolves the system's own representations** (April 2026): This principle is self-similar — it applies not just to user-contributed claims but to the system's own intermediate representations. CQ polarity assertions, chain-of-thought reasoning traces, and any \"bundled\" claim that generalizes over independently-verifiable subclaims are all \"currently undecomposed,\" not \"permanently atomic.\" A general claim should decompose into its subclaims, with QBAF strength flowing transitively from evidence on the subclaims back to the mother claim. The sorry model is recursive: each level of decomposition reveals the next level's sorry markers. (See [research/self-similar-decomposition-and-claim-ontology.md](research/self-similar-decomposition-and-claim-ontology.md))\n\n**Full exploration**: [depth.md](depth.md) — the Socratic function, universal acid, convergence thesis with MGE empirical support, Tao's verification thesis, evidence-as-subgraph, and where the truly ineffable begins.\n\n## Epistemic Humility as Architecture (Mar 28, 2026)\n\nMulti-model consensus and steelmanned critiques converge on a critical framing: Deliberus outputs are **\"mapped perspectives,\"** not objective truth. The platform makes reasoning visible, structured, and auditable — but it does NOT claim the output is correct. Every design choice embeds epistemological assumptions (the \"neutrality illusion\" critique); acknowledging this openly is both intellectually honest and strategically necessary for trust. This humility is what enables the [trust transformation](civilizational-vision.md): trust shifts from people (proxies who can be corrupted) to process (reasoning you can inspect). \"Focus on trusting the process instead of the people\" (2012). (See [research/steelmanned-critiques.md](research/steelmanned-critiques.md), [research/consensus-path-forward.md](research/consensus-path-forward.md))\n\n## Why Every Predecessor Failed — and Why This Might Not\n\n*Where the predecessors sit in the wider map of epistemic infrastructure (what exists for conclusions, quality control, agreement-shape, and procedure, and the reasoning layer that never got built) is drawn out in [The Missing Layer](the-missing-layer.md).*\n\nEvery structured argumentation platform in the graveyard (20+ since 2007) was purely left-hemisphere: decompose, categorize, score, rank. They failed because analytical tools alone don't motivate sustained use. The worldview filter — \"put on glasses you'd never choose and see what you've been missing\" — is the emotional and motivational hook that no predecessor offered. **That's not a feature. It's the soul of the product.** The analysis serves the attunement; the attunement motivates the analysis.\n\nA second failure mode, independent of the worldview gap: **every predecessor demanded formal structure as INPUT rather than generating it as OUTPUT.** Users had to manually decompose their reasoning into nodes and edges — a cognitive tax that only academics and debate enthusiasts would pay. The Lean proof assistant community solved the analogous problem (formalizing math is painful) through `sorry`-driven development: sketch the structure with holes, fill in later, let others contribute individual pieces. Deliberus applies the same pattern: free-form text or voice → LLM extraction → formal structure with `sorry` markers → community fills in the gaps. The user never needs to think formally; the formality emerges. (See [research/lean-deliberus-analogies.md](research/lean-deliberus-analogies.md))\n\n## The Combinatorial Bet (Mar 28, 2026)\n\nNo individual component of Deliberus is novel — claim extraction exists (Claimify), argument maps exist (Kialo), opinion clustering exists (Polis), epistemic scoring exists (Metaculus). What's novel is the **combination** and the **worldview-navigability layer** on top: extraction + fact/value classification + two-axis evaluation + perspective-shifting filters + progressive disclosure. Each component alone has a dead predecessor; the combination — especially with worldview lenses — creates a category that doesn't exist yet. The bridging arguments concept (detecting reasoning that's compelling across worldview groups) is a valuable analytical tool within this framework, originally synthesized during Session 3 research from Polis + argumentation theory ([research/bridging-arguments.md](research/bridging-arguments.md)).\n\n**The combinatorial bet made concrete** (Mar 31, 2026): The full pipeline is now built and deployed. Extraction → scheme detection → CQ generation → QEM gradual semantics → correction UX → agree/disagree voting → bridging feed. The bridging signal — `disagreement_factor × QBAF_strength` — is computable ONLY because Deliberus combines voting (Polis has this) with structural argument analysis (Kialo has this) with scheme-based quality assessment (nobody has this). The \"bridging\" feed mode surfaces claims where people disagree on the conclusion but the reasoning is structurally sound (CQs well-answered, green QBAF badge). This is the \"I can't refute this, which makes me reconsider\" moment — the product of the combination, invisible to any single component. *(Measured 2026-08-17: that describes the design. The shipped feed ranks by the classifier's extraction-time confidence rather than by the CQ-derived badge, so the combination is built but not yet joined at this one surface — [bridging.md](bridging.md) § The Bridging Signal.)*\n\n**First empirical validation** (Mar 31, 2026): 7 extractions (articles + PDFs) produced 318 claims, 93 cross-extraction auto-connect edges, 372 CQs, 21 contested concepts. The `@[simp]` flywheel works — each extraction enriches the graph. \"Moral urgency\" surfaced as a CRITERION contested concept (deontological vs utilitarian framing). \"Intelligence\" spans sources with different operative meanings. The combinatorial bet is producing insights no individual component would generate.\n\n**The Apr 1–15 blind-spot lesson** (Apr 15, 2026): Empirical validation is only real if *other people* can trigger it. For 15 days after Apr 1, every non-Fredrik user who tried an extraction hit a silent Pydantic validation bug that returned 0 claims — 7 target users — including a cognitive scientist and EA-community researchers — each saw Deliberus as a broken 404 on their first visit. The seed corpus (11 Mar 31 extractions, 2,130 claim nodes) was real; the adoption surface around it was invisible. Empirical validation at civilizational scale requires persistent failure observability — not just success metrics — or the pipeline will appear to work while silently losing every first user. Full incident and the forensics layer that now guards against recurrence: [research/session13-notification-snr-and-extraction-outage-rca.md](research/session13-notification-snr-and-extraction-outage-rca.md).\n\n## The Destination\n\nThe long-term vision is not a personal thinking tool — it's a **civilizational knowledge artifact**. ALL worldviews mapped onto one shared structure. Like Wikipedia but for the structure of human reasoning. The personal thinking tool is the entry point; the destination is a graph that any human can navigate through any [worldview's lens](worldview-lenses.md), at any level of complexity, to understand not just what people believe but why — and what they might be right about that you've missed.\n\nWhat Wikipedia did for facts, Deliberus does for reasoning. A **library of reasoning** — which premises lead to which conclusions, which concepts are contested and why, where evidence is strong and where it's missing. A civilizational memory: \"we already mapped this argument; here's where it's still open.\" What the scientific method did for empirical inquiry, generalized to ALL domains of human deliberation.\n\nIf this graph reaches maturity, the consequences are civilizational. [AI agents become alignable](civilizational-vision.md) — not through imposed rules but through verifying their reasoning against the natural [convergence](convergence.md) of human values at depth. Resource allocation becomes inspectable — the argumentative layer of \"why should resources flow here?\" becomes transparent, making the transition beyond pure market pricing *thinkable*. And the [bridging](bridging.md) signal — reasoning that people from opposing groups both find structurally sound — becomes computable for the first time.\n\nThe same graph is also cooperation infrastructure: an anti-rivalrous public good where every clarified premise, contested definition, evidence subgraph, and bridging argument lowers the future cost of mutual understanding. If Coase is right that efficient bargains fail because transaction costs are too high, Deliberus targets the deliberative transaction costs that keep many social conflicts artificially zero-sum — the civilizational-scale reading of [Opacity Is a Cost, Not a Mystery](#opacity-is-a-cost-not-a-mystery-aug-2026). (See [research/non-zero-sum-economics-and-civilizational-cooperation.md](research/non-zero-sum-economics-and-civilizational-cooperation.md))\n\nTerence Tao's verification thesis ([arXiv:2603.26524](https://arxiv.org/abs/2603.26524), March 29, 2026) validates this architecture from mathematics: \"AI tools should not be viewed purely through the technical lens, but through the macroscopic humanitarian lens of how our society, our shared body of knowledge and understanding, and our species benefits as a whole.\" Ideas are now cheap; verification is the bottleneck. Deliberus is verification infrastructure for human reasoning.\n\n> **Independently corroborated and bounded (2026):** a heavyweight independent research group (SciencePedia / LCoT \"Inverse Knowledge Search\", arXiv:2510.26854) reached this same premise — that knowledge systems radically compress reasoning and the lost derivational scaffolding is recoverable as a verifiable corpus — and demonstrated the reasoning-corpus + cross-model-verification + inverse-search + emergent-cross-disciplinary-graph pattern at very large scale. Their reliability mechanism is endpoint-verification-dependent, so it occupies the *verifiable-STEM* slice and structurally cannot enter Deliberus's normative / contested / no-verifiable-endpoint space — which sharpens, rather than threatens, Deliberus's distinct niche.\n\nThe strongest version of this claim: Deliberus is not a tool for alignment research — it is alignment *infrastructure*. The epistemic substrate that makes alignment possible at civilizational scale. See [research/deliberus-as-alignment-infrastructure.md](research/deliberus-as-alignment-infrastructure.md) for the full case — specific mechanisms, philosophical grounding, honest risks, and testable conditions.\n\n**Full exploration**: [civilizational-vision.md](civilizational-vision.md) — trust transformation, AI alignment, resource allocation, post-tribal epistemics, love through understanding, the evolutionary next step, Constitutional AI as snapshot vs camera.\n\n## Three Horizons (Mar 29, 2026)\n\nThe personal tool and the civilizational graph are not sequential phases with gates — they're nested. The path runs through personal utility into civilizational infrastructure, with a combinatorial demonstration in between:\n\n- **Horizon 1 (weeks)**: Personal thinking tool. \"Help me think about X.\" Paste text → structured analysis → edit/correct/decompose. Single-player utility dissolving cold-start.\n- **Horizon 2 (months)**: Combinatorial demo. The moment extraction + classification + evaluation interact and reveal something no existing tool shows. This justifies building Deliberus rather than contributing to Kialo or Polis. (See [research/combinatorial-mvp-reasoning.md](research/combinatorial-mvp-reasoning.md))\n- **Horizon 3 (years)**: The civilizational deliberation graph described above.\n\nThe Wikipedia model: you come to read, gradually start editing. Creation and exploration are the same act. Each horizon is an entry point into the next. (See [research/session7-deep-review-and-strategic-assessment.md](research/session7-deep-review-and-strategic-assessment.md) §Three Horizons)\n\n## The Dialectic Is Fractal (Mar 29, 2026)\n\nThe analysis ↔ attunement dialectic isn't a theme — it's the project's structure at every level. Philosophy (decompose vs. inhabit), strategy (civilizational graph vs. personal tool), architecture (graph nodes vs. embedding space), UX (claim structure vs. free text), data model (explicit classification vs. implicit clustering). Every design question resolves to the same dialectic. Neither pole should win. Full table in [research/session7-deep-review-and-strategic-assessment.md](research/session7-deep-review-and-strategic-assessment.md) §The Dialectic Is Fractal.\n\n## Key Tensions\n\n- **Analysis vs attunement**: The core dialectic (see above). Both must coexist; neither should dominate\n- **Internalization vs collective compute**: How much should users internalize reasoning skills vs. let the system compute for them? Both, but the balance matters\n- **Structure vs friction**: Structured argumentation is effortful. LLMs must reduce the friction enough that structuring becomes viable at scale. The resolution: progressive complexification — simple surface, deep structure available on demand. The Lean analogy deepens this: Lean's `sorry` mechanism lets users sketch proof *structure* with explicit gaps, and others fill in the gaps later. For Deliberus, free-form text or voice input → LLM extraction → argument structure with `sorry` markers where reasoning is incomplete. Contribution barrier drops to zero while structural rigor is maintained underneath. (See [research/lean-deliberus-analogies.md §1](research/lean-deliberus-analogies.md))\n- **Formal logic vs defeasible reasoning**: Lean-style verification is too rigid for human arguments. Need defeasible argumentation logic while keeping the \"proposer + verifier\" architectural bones. The kernel should be small and auditable (Lean's is ~5,000 lines of C++); it checks argument *structure*, not truth. The community judges meaning. (See [research/lean-deliberus-analogies.md §3](research/lean-deliberus-analogies.md))\n- **Individual brilliance vs collective intelligence**: The system must be designed so that the COLLECTIVE output is wiser than any individual participant — including the platform's designers. (See [conceptual-threads.md §Thread 2: Hybrid Intelligence](conceptual-threads.md))\n- **Personal tool vs societal infrastructure**: The path to societal infrastructure runs through personal utility. Single-player first, social features emerge from shared artifacts. But the destination is civilizational, not personal. (See [research/single-player-utility.md](research/single-player-utility.md))\n\n## Academic Alignment (Mar 31, 2026)\n\n**\"Argumentative Human-AI Decision-Making\"** (Vasileiou, Rago, Toni, Yeoh — AAMAS 2026 Blue Sky) states Deliberus's thesis in academic form: \"AI agents should engage in dialectical processes where decisions are contestable and revisable.\" Francesca Toni's group at Imperial College London (ArgLLM-App, QEM collaborator Potyka) is the closest academic alignment. Deliberus goes significantly beyond their ArgLLM-App: extraction pipeline + scheme detection + CQ generation + QEM + correction UX + Temporal workflows vs binary decision support. The \"bridging argument\" concept (extending bridging from statements to reasoning structure) remains genuinely novel — no existing system computes bridging metrics over argument graphs. Blair et al. (\"The Structure of Bridging\") solves the formal bridging problem with partition-agnostic pairwise disagreement; Deliberus extends this to reasoning quality, not just statement agreement. (See [research/unbuilt-features-research-mar31.md](research/unbuilt-features-research-mar31.md), [research/qbaf-gradual-semantics-research.md](research/qbaf-gradual-semantics-research.md))\n\n**March 2026 — a convergence moment.** Five or more papers on AI-augmented deliberation published in a single month: DCI (\"From Debate to Deliberation,\" Prakash, arXiv:2603.11781) validates that structured deliberation with typed epistemic acts significantly outperforms unstructured debate — its 14 typed acts closely mirror Walton's scheme classification. ARGSBASE (EACL 2026) demonstrates multi-agent structured human-AI deliberation. \"Argumentative Human-AI Decision-Making\" (arXiv:2603.15946) calls for AI that reasons WITH humans, not FOR them. And Terence Tao's philosophical essay \"Mathematical Methods and Human Thought in the Age of AI\" ([arXiv:2603.26524](https://arxiv.org/abs/2603.26524), March 29) articulates the humanitarian lens: AI should augment, not replace; verification infrastructure (Lean for math, structured argumentation for reasoning) is the critical bottleneck; and meaning remains irreducibly human. The academic world is converging on the exact space Deliberus occupies — but nobody builds persistent, evolving argument graphs with scheme classification and critical question generation. Full analysis: [research/ai-safety-and-tao-augmentation-research.md](research/ai-safety-and-tao-augmentation-research.md).\n\n## Adjacent Motivations\n\n- **AI Safety / AI-Enhanced Decision-Making**: 80,000 Hours identifies AI-enhanced decision-making as a high-impact, *highly neglected* problem area — \"only a handful of projects\" building these tools, needing \"a few hundred more thoughtful, entrepreneurial people.\" They call for epistemic tools (fact-checkers, forecasting) and coordination tools (negotiation, transparency) but **do not mention structured argumentation or deliberation platforms** — the exact gap Deliberus fills. Deliberus is the missing *structured reasoning layer* connecting facts to decisions through transparent argument analysis. \"Tools arriving just a few months earlier could mean critical safety commitments get nailed down before we develop dangerous AI systems.\" (See [research/80000-hours-ai-decision-making.md](research/80000-hours-ai-decision-making.md))\n- **Post-scarcity vision**: Contributing to a civilization that moves beyond cutthroat competition, with healthier economic underpinnings.\n- **Digital deliberative democracy**: Consistent thread from CRM outreach to youth parties (CUF, Ung Pirat, Gröna Studenter) through to AI safety communities.\n\n## Key Quotes\n\n> \"I'm an EA but also anticapitalist in the long run, but in order to figure out (and transition into) something better than capitalism, we need to build a collaborative sensemaking platform, Deliberus or something like it, a way for humanity to mindmeld with itself, to deeply and thoroughly process all knowledge, all beliefs and the evidence supporting it, and so distill our values to their essence, at least represented in language.\"\n\n> \"[Science] is a subset of consensus tech but with limited application so that something like this would have more impact than science.\"\n> — Curtis SerVaas, Thiel Fellowship application (2012)\n\n> \"Basically the approach I would think is most fruitful is looking at psychological principles and considering them microforces that drive part of the machinery of this pseudo-AGI, and seeing how you can focus and guide them using checks and balances in the design of the app(s) involved.\"\n> — Fredrik (Jan 2012, conversation with Curtis SerVaas)\n\n> \"Focus on trusting the process instead of the people.\"\n> — Fredrik (2012, article outline). The earliest formulation of trustless collaboration — the process (structured argumentation with machine-checked validity) is trustworthy even when the people are unknown. See [research/simplenote-archive-analysis.md §5](research/simplenote-archive-analysis.md)\n\n> \"Self-governance has never been so feasible and at the same time crucial as at this point in human history.\"\n> — Fredrik (2012, Facebook post linked from article outline)\n\n---\n\n**See also**: [Analysis ↔ Attunement](analysis-and-attunement.md) · [Worldview Lenses](worldview-lenses.md) · [Depth](depth.md) · [Bridging](bridging.md) · [Convergence](convergence.md) · [Civilizational Vision](civilizational-vision.md) · [History](history.md) · [Conceptual Threads](conceptual-threads.md) · [UX Principles](ux-principles.md)\n"}