{"path":"research/simplenote-archive-analysis.md","content":"# Simplenote Archive Analysis — Earliest Deliberus Notes (2011-2012)\n\n**Date**: March 28, 2026 (Session 4)\n**Source**: FERMI drive, Simplenote Support Notes\n**Documents analyzed**: 7 Simplenote text files from the genesis period of Deliberus\n\nThese notes predate the Meteor prototype, the Nils Janse partnership, and most of the chat logs. They represent the earliest surviving written formulations of the Deliberus vision, from Fredrik's time at Leverage Research and concurrent conversations with Ivan Phillips, Tom Boniecki, Ben Hyink, Curtis SerVaas, and the LessWrong community.\n\n---\n\n## 1. Leverage.txt — Working Notes from Leverage Research (~2012)\n\nNotes from Fredrik's ~3-week stint at Leverage Research, Berkeley. Working through the design space of \"consensus tech.\"\n\n### Verbatim Content (key sections)\n\n**Why do people disagree — main sources of disagreements:**\n\n> Investigations too short, too limited, too weak, tools aren't very powerful\n\n> **Coarse-grained discourse**\n>     Improving needs to move in the direction of finer grains?\n>         **If radically improved, could be likened to the advent of language itself**\n>     Allows for integration and synchronization at a more detailed level\n\n> Missing preconditions\n>     psychological\n>         honestly seeking the truth, be okay with radical change\n>         being open to having one's views refined and challenged fully\n>     technical\n>         features missing, like a car or airplane\n\n> Sectors\n>     Prescriptive, normative, descriptive\n>         Prescriptive\n>             very sensitive to\n>             **start with opaque axioms, drill down carefully with time**\n\n> What makes people agree? What is agreement?\n>     **Isomorphisms between paths to conclusion?\n>     Or just isomorphism between conclusions?**\n>\n>     **Illusory agreement common/pervasive?**\n\n### Analysis\n\n**\"Coarse-grained discourse\" as civilizational framing**: The claim that improving discourse granularity \"could be likened to the advent of language itself\" is the strongest framing of Deliberus's mission anywhere in the project. Language enabled fine-grained communication between minds; Deliberus enables fine-grained *reasoning* between minds. This is not \"a better debate tool\" — it's a new medium for collective cognition. This framing is not in [vision.md](../vision.md) and deserves elevation.\n\n**Disagreement taxonomy**: A structured breakdown of WHY people disagree that isn't captured elsewhere:\n1. **Investigations too limited** — tools aren't powerful enough for the complexity\n2. **Coarse-grained discourse** — language is too imprecise for the nuance needed\n3. **Missing psychological preconditions** — people aren't honestly seeking truth\n4. **Missing technical preconditions** — the infrastructure doesn't exist yet (\"like a car or airplane\")\n5. **Sectoral differences** — prescriptive/normative/descriptive, academic/corporate/government, local vs global objectives\n\n**\"Illusory agreement\"**: The question \"Isomorphisms between paths to conclusion? Or just isomorphism between conclusions?\" is the Emanuel discourse-layer insight (2013, see [voice-memo-emanuel-sofia.md](voice-memo-emanuel-sofia.md)) formulated *two years earlier* and more precisely. Two people can share a conclusion via completely different reasoning paths — and this masks real disagreement that surfaces later under different conditions. The Swedish \"barnfattigdom\" debate was exactly this: agreement on the conclusion (children shouldn't be poor) masking disagreement on definitions and mechanisms.\n\n**\"Start with opaque axioms, drill down carefully with time\"**: The earliest surviving formulation of the \"No Copout Axioms\" principle (see [vision.md §The \"No Copout Axioms\" Conviction](../vision.md)). Already in 2012, the design intent was to start with what LOOKS like bedrock and progressively decompose it.\n\n---\n\n## 2. Priorities for consensus tech.txt — The Priority Ordering\n\nComplete verbatim content:\n\n> Priorities for consensus tech\n>\n> 1. Intelligence amplification for groups of ~2-10 persons\n> 2. Massively distributed decision-making/deliberation\n> 3. Critical thinking/logic education\n\n### Analysis\n\nThree lines, but architecturally significant. This is an explicit **build ordering** not captured in existing docs:\n\n- **Priority 1** (small groups): Aligns with the single-player + small-group entry point from [single-player-utility.md](single-player-utility.md). Start with \"help me think\" for one person, extend to 2-10 for team decisions.\n- **Priority 2** (massive scale): The civilizational deliberation graph from [vision.md §The Destination](../vision.md).\n- **Priority 3** (education): The Kialo beachhead analyzed in [adoption-problem.md](adoption-problem.md) — education works but may be a ceiling.\n\nThe ordering says: build the thinking tool first, scale it second, education is a byproduct not the primary goal.\n\n---\n\n## 3. Ivan Phillips note.txt — \"A Tool to Industrialize Knowledge\" (Jan 12, 2011)\n\nThe most important document in this archive. A Facebook note by Ivan Phillips with an extended comment thread. History.md references Ivan for the \"hybrid AGI\" quote (\"a system of argument map + human society might be a hybrid AGI\"), but this is the FULL original text — a complete design brief for what would become Deliberus.\n\n### Ivan's Note (verbatim, key passages)\n\n> The sphere of human knowledge is growing very quickly. [...] it's already outgrown the capacity of our poor monkey brains to handle it. 100 years ago, a single physicist might have been able to understand all of physics. Today, a physicist would have a hard time understanding a small sub-domain of physics.\n\n> **What we need is a system in which each page is a proposition, and each link between pages is an inference.** The system would encode deductive inferences and inductive ones, and likely more than one flavor of inductive inference. The goal is to represent what we know and why we know it.\n\n> Even if you and I disagree on a particular conclusion, **I think we ought to be able to agree that *if* the premises are true, and *if* certain forms of inference hold, *then* the conclusion would be true.**\n\n> [On gaming:] some governments, corporations and special interests will want to game the system, and strategies for dealing with this will need to be developed.\n\n> Ultimately, I think that there will be **entire industries focused on getting information into this system (systems like this one) and getting information out of it.** It just seems inevitable that the volume of human information will continue to grow rapidly, and the capacity of our minds won't be able to keep up.\n\n### The Comment Thread\n\n**Tom Boniecki** (Jan 12, 2011):\n\n> I think the idea is not to write a program to understand the data and draw logical conclusions but merely assist in organizing the knowledge and keep humans on a straight and narrow by making things clearer and providing templates and reminding of common mistakes. **I see it as a spreadsheet for knowledge rather than an expert system.**\n\n> By adjusting weights and confidences you will be able to come to different conclusions but you will have to **state those confidences explicitly** instead of making leaps of logic and in the process maybe it will become clear that your reasoning or conclusions are not substantiated.\n\n> **The system should also allow the users to argue (expose) semantics.** In a proposition \"The Bears are the best\" a lot hinges on the definition of \"the best\". Even The Bears could have multiple definitions.\n\n> Fredrik, many arguments are strictly about semantics and I think people are used to arguing them. I think **being organic is not necessarily a good thing. It could make people rely on their old organic instincts** and make them use tactics they use in face to face arguments.\n\n**Ben Hyink** (Jan 12, 2011):\n\n> [The system] would be very useful for governments and organizations to use when setting policy, and useful for researchers who want to identify weak parts of our inference or new territory for research projects.\n\n> [The system] would work best in dealing with established knowledge, scientific methodology, and empirically testable theories, but would be **limited in making useful comparisons between systems of thought with radically different and widely varying priors** (often expressed in human-unfriendly let alone machine-unfriendly ways), like the most abstract \"philosophy\" level of any discipline [...] having a philosophy badly mismatched to the game being played can make one completely uncompetitive.\n\n> Maybe, over a sufficiently long time with intensive data entry made by a widespread collaboration of philosophers, some headway might be made into at least **defining all the priors being posited about some topic** and claims about the rational inferences and deductions that can be made from them (priors of epistemology would be at the root, btw).\n\n**Fredrik Bränström** (Jan 12-17, 2011):\n\n> [On semantics:] I'd like to include you both in a brainstorming session over Skype sometime in the next few days. (I want to build something like this during the next few months, and so I stand before some icky questions and decisions.) Adam Safron, famous neuroscientist, will also be joining in.\n\n> **Perhaps an invite-only prediction market would be a good model, where people are awarded for making contributions to what later turns out to be well-formed argumentation?** At least to start with.\n\n> Tom, organic as in corresponding well to how brains actually process/store meaning. We'll make it as hard for people as possible to revert to their old ugly tactics and rhetorics — the UI should help (**splitting compound statements and letting other people judge each part of the argument individually**, for example), and in general **the constraining data structure should make invalid/ungrounded inference painfully obvious**, especially with constant feedback/tweaking of statements/voting from other users.\n\n> [Shared Tim van Gelder research]: \"In the semester in which we used automated argument mapping exercises, we measured an improvement of **0.45 standard deviations** in critical thinking skills. [...] the critical thinking gains measured during our 12 week course are close to that which could be expected to result from **three years of undergraduate education**.\"\n\n### Analysis\n\n**Ivan's note is the AIF in plain English** — \"each page is a proposition, each link is an inference\" is exactly the I-node/S-node model from Chesnevar et al.'s Argument Interchange Format (2007, see [academic-foundations.md §6](../academic-foundations.md)), arrived at independently.\n\n**The kernel separation principle**: \"if the premises are true AND the inferences hold, THEN the conclusion is true\" is the separation of structural validity from content truth that Lean's kernel embodies (see [lean-deliberus-analogies.md §3](lean-deliberus-analogies.md)). Ivan articulated it in 2011.\n\n**Tom's \"spreadsheet for knowledge\"** is a better analogy than \"Wikipedia for arguments.\" A spreadsheet doesn't tell you what to put in the cells; it gives you a structure for organizing whatever you put in. It makes you state things explicitly. It computes consequences. Deliberus is a spreadsheet for reasoning.\n\n**Tom's semantics warning** — \"being organic is not necessarily a good thing\" — is an early formulation of what became the interpassivity warning (P14 in [ux-principles.md](../ux-principles.md)): don't let people fall back on intuitive rhetorical habits; the structure should force explicitness.\n\n**Fredrik's prediction market idea** (Jan 2011) — \"awarded for making contributions to what later turns out to be well-formed argumentation\" — is the epistemic gamification/calibration concept, articulated four years before Metaculus launched. This connects directly to the calibration scoring system in [epistemic-gamification.md](epistemic-gamification.md).\n\n**Fredrik's \"splitting compound statements\"** — is atomic claim extraction (Claimify 2025, Microsoft Research), described in 2011.\n\n**Ben's \"radically different priors\"** — flagged the fact/value boundary ([conceptual-threads.md §Thread 3](../conceptual-threads.md)) and the limits of formal systems when dealing with incommensurable worldviews.\n\n---\n\n## 4. Article draft.txt — Short Article Draft (~2012)\n\n### Verbatim Content\n\n> We need new frameworks to nurture our collective intelligence. Social media in its current instantiation is a playground at best.\n\n> The race now is to leverage and structure collaboration. We need to distribute the tasks of knowledge creation and analysis so that we each can contribute with local/special/deep information, figure out how to let people prune and refine the data in a stable manner, visualize the results pedagogically and intuitively to feed more intelligence back to the users. The tricky bits are going to be designing checks and balances appropriate to optimize logical coherence in the process, as well as dealing with different motivations in users.\n\n> **A milestone will be reached when the users of the descendant systems of today's \"social media\" cross the threshold of having more influence over the design of the systems than do the companies or institutions that constructed them.** Will the result be more bottom-up than top-down? What will be the optimal balance between the two?\n\n### Analysis\n\nThe self-governance milestone — users eventually have more influence over system design than the builders — is not captured elsewhere. It connects to Bluesky's algorithmic marketplace ([feed-algorithm-design.md](feed-algorithm-design.md)) and to the \"kernel should be deliberatable\" insight from the Lean analogy ([lean-deliberus-analogies.md §3](lean-deliberus-analogies.md)).\n\n---\n\n## 5. Article.txt — Full Article Outline (~2012)\n\nThe richest single document. An article outline covering the full Deliberus vision with references.\n\n### Key Verbatim Passages\n\n> **Focus on trusting the process instead of the people?**\n\n> With the help of the frontal cortex of millions of other humans, we can make inferences that we don't have time to bear out on our own — time, setup cost, distraction cost, energy for thinking deeply and rationally. **Personal strategies to avoid bias are more costly than strategies embedded in the tools we use.**\n\n> Provide people with a framework to **inspect reasoning together, how it intersects and overlaps, instead of working against one another.** Move from politics and debate as a largely competitive arena to a synergistic, collaborative platform.\n\n> Yochai Benkler talks about Social Production; it should be harnessed for the **highest meta-good of rationality** — the methods that helps us think about the real world and define and achieve our goals within it.\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\n> **\"Self-governance has never been so feasible and at the same time crucial as at this point in human history.\"**\n\n> Evolutionary shifts should be mentioned [...] **evolutionary biology suggests next big thing is collaboration on a higher level of organization. We've seen this again and again, with the birth of multicellular life, culture, civilization, etc.** This is why we have something called morality it seems; we are shifting to a collaborative default as a species, but it looks like we need a better framework for it, or maybe just tighter communication.\n\n> Cognitive surplus — enzyme folding gamers, passive TV-watchers of 90's now awakening to collaborate [Clay Shirky reference]\n\n> Nobody reads null-finding journals. Tough to keep up with information accrual, integration with earlier theories/hypotheses, hard to be wise in a fast-moving world. Clarify reasoning to expose for easier evaluation, **perhaps in time entirely replace peer-review publishing.**\n\n### Analysis\n\n**\"Focus on trusting the process instead of the people\"** — This is trustless collaboration as a design principle, long before we derived it from Lean's `sorry` pattern in Session 4. The process (structured argumentation with machine-checked validity) is trustworthy even when the people are unknown. This exact phrasing connects to [lean-deliberus-analogies.md §5](lean-deliberus-analogies.md) (trustless collaboration through structural verification).\n\n**The earliest vision statement** — \"an algorithm for humanity to reason and make decisions together... combining the democratic elements and general efficiency of a decentralized system like markets with the discerning force of logic\" — is the core mission in its purest, earliest form. It belongs alongside the later formulations in [vision.md §Key Formulations](../vision.md).\n\n**The evolutionary biology frame** — Deliberus as the next evolutionary step in collaboration, following multicellular life → culture → civilization. This is not in any existing doc and provides a powerful narrative frame for the civilizational ambition.\n\n**\"Personal strategies to avoid bias are more costly than strategies embedded in the tools we use\"** — This phrase appears nearly verbatim in [vision.md §The Problem](../vision.md) (\"Personal strategies for rationality are more costly and fragile than strategies that we can build into tools that we use collaboratively\"). The Simplenote note is the original source.\n\n**Yochai Benkler / Clay Shirky** references establish the intellectual lineage: social production (Benkler) + cognitive surplus (Shirky) → applied to the \"highest meta-good of rationality.\" This lineage isn't captured in existing docs.\n\n---\n\n## 6. Summaries.txt — Feature Brainstorming + Facebook Group References\n\nWorking notes linking to Facebook group discussions with feature ideas and design decisions.\n\n### Key Verbatim Passages\n\n> General idea; recursive argmap\n>     Iterative, collaborative refinement\n>     Crowdsourcing + automation, language parsing\n>     Microtasks, or just surfing around editing\n>     **Queues or feeds of needy args/claims**\n\n> User influence (lots of comments by James and Alex):\n>     \"might be useful to write out the whole gamut of possible privileges and arrange them by 1. frequency needed by the site and 2. responsibility needed by user.\"\n>     **\"When a normal user's validity votes overlap with the judgments of validators to a certain extent (9/10 maybe) after a certain period of time, they will achieve the validator title, either by validator selection or automatically...\"**\n>     \"the criteria for selection to a higher user rank (as currently imagined) is merely logic fu, which is universal and not dependent on the topic.\"\n>     \"I like the idea of **noting discrepancies between user votes and sub-admin votes.** You could perform automated validation of the validators by analyzing their results on the standard deviation with other users and other validators\"\n>     \"There could just be a queue of all validations/actions that have been performed at each level, higher-ups would watch the queue, **promote where recommended by the system** (where the stats on actions performed look best; the system sorts the users to suggest which to promote to the next level).\"\n\n> **De Bono thinking hats as funnels for crowdsourcing changes — a yellow queue, a black one, red, blue**\n> Which queues/tasks are best adapted to a certain thinking hat?\n> How to adapt the UI and guide the user in the best way for a certain task?\n\n### Analysis\n\n**\"Queues of needy args/claims\"** — Evidence gap detection as a feed design principle, from 2012. This is exactly the \"evidence gaps are contribution invitations\" principle (P17 / feed-algorithm-design.md §2.5), conceived 14 years before we formalized it.\n\n**The user permission hierarchy** is more detailed than Shneiderman's Reader→Leader model ([academic-foundations.md §8](../academic-foundations.md)):\n- Normal user → validator → inspector → admin\n- Promotion is **automated by the system** based on vote-accuracy metrics\n- \"Logic fu is universal and not dependent on the topic\" — domain-independent epistemic competence\n- **Discrepancy detection** between user votes and validator votes — this is cross-adversarial quality rating detection, independently conceived in 2012\n\n**De Bono thinking hats as contribution funnels** — This is a genuinely novel UX concept not in any existing doc. Instead of a generic \"contribute\" button, route users into contribution MODES based on what type of thinking the argument needs:\n- **Yellow hat queue** (optimistic): \"What's good about this argument? What supports it?\"\n- **Black hat queue** (critical): \"What could go wrong? What attacks this?\"\n- **Red hat queue** (emotional): \"How does this make you feel? What's your gut reaction?\"\n- **Blue hat queue** (meta): \"Is this argument structured correctly? What's missing?\"\n- **Green hat queue** (creative): \"What alternatives haven't been considered?\"\n\nThis connects to the pre-categorized microinteractions in [mobile-argument-ux.md §2.3](mobile-argument-ux.md) and to the `sorry` model: each hat mode produces a different TYPE of contribution that fills in different gaps.\n\n---\n\n## 7. D todo.txt — Development Todo (~2012)\n\nShort development notes:\n\n> Versions: attributes? CTI?\n> Tags - on links or claims?\n> copremises - be able to move them? or just CRUD? reset stats when moving?\n>\n> Recruit Morendil: http://lesswrong.com/lw/1qq/debate_tools_an_experience_report/\n> Recruit SU philosophy people? http://www.philosophy.su.se/personal/salwen.htm\n> NLP people (Curtis knew someone through LW): http://groups.inf.ed.ac.uk/probmodels/people\n\n### Analysis\n\nHistorically interesting: recruitment targets included LessWrong user Morendil (who had written about debate tools), Stockholm University philosophy faculty, and NLP researchers via Curtis's LW connections. The \"copremises\" question (can you move them? reset stats?) shows early thinking about the atomic claim manipulation UX.\n\n---\n\n## What's Genuinely New (Not in Any Existing Doc)\n\n| Finding | Source | Connection to Existing Work |\n|---------|--------|-----------------------------|\n| \"Coarse → fine-grained discourse = advent of language\" | Leverage.txt | Elevates the mission beyond \"better tool\" to \"new medium.\" See [vision.md](../vision.md) |\n| \"Illusory agreement\" — path vs conclusion isomorphism | Leverage.txt | Precedes Emanuel (2013) by 2 years. See [conceptual-threads.md §Thread 4](../conceptual-threads.md) |\n| Disagreement taxonomy (5 sources) | Leverage.txt | Structured WHY not in any doc |\n| Priority ordering: small groups → distributed → education | Priorities.txt | Explicit staging. See [single-player-utility.md](single-player-utility.md) |\n| Ivan's full note + comment thread | Ivan Phillips.txt | AIF independently; kernel separation; \"spreadsheet for knowledge\" |\n| Tom Boniecki's \"spreadsheet for knowledge\" analogy | Ivan Phillips.txt | Better than \"Wikipedia for arguments\" |\n| Fredrik's 2011 prediction-market + reputation idea | Ivan Phillips.txt | Epistemic gamification, 4 years pre-Metaculus. See [epistemic-gamification.md](epistemic-gamification.md) |\n| \"Splitting compound statements\" = atomic claim extraction | Ivan Phillips.txt | Claimify (2025), described in 2011 |\n| \"Focus on trusting the process instead of the people\" | Article.txt | Trustless collaboration. See [lean-deliberus-analogies.md §5](lean-deliberus-analogies.md) |\n| \"Algorithm for humanity to reason together\" | Article.txt | Purest early vision statement |\n| Evolutionary biology frame (multicellular → civilization → this) | Article.txt | Vision framing for civilizational ambition |\n| Yochai Benkler / Clay Shirky intellectual lineage | Article.txt | Social production + cognitive surplus → rationality |\n| Self-governance milestone (\"users > builders\") | Article draft.txt | Community governance. See [feed-algorithm-design.md](feed-algorithm-design.md) |\n| De Bono thinking hats as contribution funnels | Summaries.txt | Novel UX concept. See [ux-principles.md](../ux-principles.md) |\n| \"Queues of needy args/claims\" | Summaries.txt | Evidence gap detection, 2012. See [feed-algorithm-design.md](feed-algorithm-design.md) |\n| Automated user promotion via vote-accuracy | Summaries.txt | See [epistemic-gamification.md](epistemic-gamification.md) |\n| Discrepancy detection between user/validator votes | Summaries.txt | Cross-adversarial quality rating, independently conceived |\n\n---\n\n## Cross-References\n\n- **History**: [history.md](../history.md) — Ivan Phillips, Tom Boniecki, Ben Hyink, Leverage Research period\n- **Vision**: [vision.md](../vision.md) — earliest formulations, evolutionary frame, coarse-grained discourse\n- **UX**: [ux-principles.md](../ux-principles.md) — De Bono hats, \"spreadsheet for knowledge,\" contribution queues\n- **Conceptual threads**: [conceptual-threads.md](../conceptual-threads.md) — illusory agreement (Thread 4), trustless collaboration (Thread 7)\n- **Related primary sources**: [chat-logs-analysis.md](chat-logs-analysis.md) (2010-2014 collaborator conversations), [voice-memo-emanuel-sofia.md](voice-memo-emanuel-sofia.md) (2013 discourse layer), [meteor-prototype-analysis.md](meteor-prototype-analysis.md) (2013-2014 prototype)\n\n---\n\n## Source Files\n\n| File | Location |\n|------|----------|\n| Leverage.txt | FERMI drive, Simplenote Support Notes |\n| Priorities for consensus tech.txt | same |\n| Ivan Phillips note.txt | same |\n| Article draft.txt | same |\n| Article.txt | same |\n| Summaries.txt | same |\n| D todo.txt | same |\n"}