# Aquin Strategic Vision **Project**: slkm_bundle (Aquin) **Product name**: Aquin **Document status**: Candidate canonical living vision **Last updated**: 2026-06-27 **Owner**: Ryan **Revision focus**: Incorporates the strategic implications of the current graph, conflict-agent, Hamming-window, rate-distortion, science-ledger, entropy-audit, sum-over-histories, trail-compression, and disputatio wireframes. Aquin is a native, local-first knowledge workbench for authoring, refining, reviewing, testing, and exporting high-fidelity OKF knowledge artifacts. It is built around one contract: knowledge should remain human-owned, provenance-rich, inspectable, portable, and useful to both people and agents. Aquin should not become another note app, chat wrapper, citation manager, or productivity dashboard. Its strategic identity is sharper: Aquin is an epistemic operating environment for transforming sources, questions, claims, links, doubts, and judgments into trustworthy knowledge artifacts. SLKM stands for Statistical Learning Knowledge Map. In the current product language, Aquin is the Mac application and workspace; OKF is the portable knowledge format; SLKM remains the conceptual foundation for bundles, refinement, typed relationships, and high-signal knowledge mapping. ## One-Sentence Vision Aquin helps people convert messy inquiry into portable, provenance-rich knowledge artifacts that preserve truth, improve judgment, and compound toward human flourishing. ## Recommended Vision Statement Aquin exists to help serious knowledge workers, researchers, builders, teachers, and organizations pursue truth with better tools. It gives them a local-first workspace where sources remain traceable, claims can be challenged, reasoning paths can be compared, uncertainty can be audited, and human judgment remains explicit. The result is not merely a larger collection of notes. The result is a trustworthy OKF artifact that another person or agent can inspect, reuse, extend, and act on. The moral center of Aquin is the ordered pursuit of truth for the common good. AI should not make knowledge less accountable. It should make claims easier to inspect, sources easier to verify, disagreements easier to reason through, and good decisions easier to preserve. Aquin is designed to help users create knowledge that serves people beyond themselves. ## Candidate Short Taglines - **Truthful knowledge, portable by design.** - **A local-first workbench for accountable knowledge.** - **From sources to judgment to trustworthy artifacts.** - **A Memex for the age of agents.** - **Human judgment, agent acceleration, provenance everywhere.** - **Knowledge work that compounds toward human flourishing.** ## Naming And Branding The product is called **Aquin** for now. Use these terms consistently: - **Aquin**: the native local-first Mac application and workspace. - **OKF project**: a folder-backed project package containing canonical Markdown/YAML knowledge files and project-local Aquin state. - **OKF artifact**: a refined, portable knowledge output intended for humans, agents, or both. - **SLKM**: the foundational knowledge-map philosophy behind high-fidelity, provenance-rich bundle work. - **Aquin Agent**: a bounded assistant that helps navigate, diagnose, propose, or review work without silently mutating canonical knowledge. - **Research thread**: a durable line of inquiry connecting sources, claims, hypotheses, judgments, and outputs. - **Judgment gate**: an explicit human decision point where the user accepts, rejects, incorporates, refutes, signs off, or records rationale. - **Cumulative Science Ledger**: the future-facing lineage layer that records how research threads supersede, extend, replicate, vary, or deprecate one another. Future naming can evolve, but the architecture must continue to protect the central contract: ordinary files are canonical, projections are rebuildable, and human judgment remains visible. ## Core Vision Aquin helps users create high-quality OKF knowledge artifacts from authored notes, ingested sources, graph analysis, structured review workflows, and local AI assistance. The product exists for outward-facing knowledge work. It helps a user, team, organization, or community create artifacts that another person or agent can trust and use. Personal learning can happen along the way, but the primary success measure is the quality, clarity, provenance, auditability, and portability of the resulting OKF artifacts. The durable source of truth is human-readable OKF Markdown with YAML frontmatter. Aquin may create graph projections, SQLite mirrors, review ledgers, saved views, local model suggestions, research threads, entropy diagnostics, and export readiness checks, but those are supporting structures. They must never obscure or replace the canonical files. Aquin's distinctive promise is that it treats knowledge as something that can be authored, tested, challenged, compressed, audited, exported, and improved over time. It is not satisfied with capture. It must help the user move from capture to clarity, from clarity to judgment, and from judgment to trustworthy transmission. ## Why This Exists Aquin began from practical machine learning and AI engineering work: reading complex material, preserving source fidelity, extracting important structure, and turning that work into usable knowledge. Its deeper motivation is the ordered pursuit of truth and the common good. Technology should help people think more clearly, preserve what is true, and act with better judgment. In an era of abundant AI-generated and AI-transformed content, the world needs tools that help people produce high-signal knowledge rather than more noise. Aquin is a response to that need. The system exists to: - Give users sovereign local-first control over their knowledge projects. - Preserve source fidelity and provenance through ingest, editing, review, export, and later revision. - Support direct human authorship in ordinary Markdown/YAML files. - Route assisted, uncertain, adversarial, or generated changes through explicit human review. - Use graphs and research workflows as rebuildable analysis tools. - Help users focus on important, high-leverage problems instead of merely convenient tasks. - Help users allocate scarce human attention where judgment, creativity, and moral responsibility matter most. - Expose uncertainty, missing citations, stale claims, contradictions, and unsupported leaps. - Record decisions and rationales so future users can see not only what was concluded, but why. - Produce OKF artifacts that are readable by humans and structured enough for agents. - Keep the final artifact portable, inspectable, and useful outside Aquin. ## Strategic Thesis The world does not mainly need faster note-taking. It needs better truth-preserving systems for turning information into accountable knowledge. Aquin should therefore optimize for five outcomes: 1. **Truthfulness**: Claims should be traceable, challengeable, and reviewable. 2. **Judgment**: Important decisions should remain human, explicit, and recorded. 3. **Portability**: Knowledge should survive outside the application that produced it. 4. **Leverage**: Agents should reduce drudgery while increasing accountability. 5. **Flourishing**: The product should help people use knowledge in service of better decisions, better institutions, better education, and better work for others. This means Aquin should resist the common failure mode of AI tools: replacing hard thinking with polished ambiguity. Aquin should instead make the hard parts visible: sources, conflicts, uncertainty, lineage, decisions, tradeoffs, and the limits of the current model. ## Design Ethos From The Current Wireframes The attached wireframes point to a stronger product identity than a generic local-first knowledge manager. Together they imply an error-correcting, judgment-centered knowledge system. ### 1. Graph Visualization: Knowledge As Navigable Structure The graph visualization expresses the principle that knowledge is not just stored text. It is a network of principles, evidence, roles, practices, outcomes, provenance, and typed relationships. Strategic implications: - Aquin should make relationships first-class, not decorative. - Nodes should carry roles such as principle, evidence, source, claim, practice, outcome, decision, actor, and open question. - Edges should use meaningful verbs, not vague links. - Provenance, role filters, timelines, and saved lenses should help users inspect why a node matters. - The Aquin Agent should guide exploration through visible graph paths rather than hidden chat-only reasoning. ### 2. Conflict Agent And Error Injection: Knowledge Must Survive Adversarial Pressure The conflict-agent wireframe introduces a provocative mode where plausible-but-false claims are injected, then routed through refutation or incorporation workflows with a human judgment gate. Strategic implications: - Aquin should deliberately test knowledge artifacts against plausible error, not merely summarize them. - Contradiction, objection, counterexample, and uncertainty should be productive states. - The user should be able to refute a claim, incorporate it as a hypothesis, or mark it as unresolved. - Important decisions should require rationale, especially when the system is handling ambiguity or contested claims. - The review ledger should preserve how the model was strengthened by conflict. ### 3. Hamming Important Problem Window: Focus Where Importance And Leverage Intersect The Hamming-window wireframe gives Aquin an explicit attention ethic: daily work should concentrate where importance to the field or humanity intersects with current tooling leverage. Strategic implications: - Aquin should help users choose better work, not merely organize existing work. - Importance and leverage should be visible dimensions in planning. - Low-importance, high-leverage work should be treated as a seductive distraction. - Important, low-leverage work should remain visible but may require better tools, collaborators, or timing. - Important, high-leverage work should become the default daily focus. ### 4. Rate-Distortion Allocator: Human Attention Is Scarce The rate-distortion allocator frames knowledge work as an allocation problem between agent time and human attention, guided by current knowledge gaps and expected information gain. Strategic implications: - Aquin should distinguish work that agents can safely execute from work that needs human judgment. - Human attention should be reserved for creativity, synthesis, prioritization, and morally significant decisions. - The system should recommend allocation based on distortion, uncertainty, work items, energy, and historical outcomes. - Agent suggestions should remain private, reviewable, and dismissible. - Aquin should help users reduce uncertainty per unit of attention, not maximize activity. ### 5. Cumulative Science Ledger: Knowledge Evolves Through Lineage The science-ledger wireframe shows research threads connected by relationships such as supersedes, extends, and replicates-with-variation. Strategic implications: - Aquin should treat knowledge as temporally evolving, not static. - Research threads should have lineage, status, owners, timestamps, and change history. - Status changes should propagate according to visible lineage rules. - Replication, variation, extension, and deprecation should be normal knowledge operations. - Public or shared artifacts should eventually carry enough lineage to be trusted beyond the originating workspace. ### 6. Shannon Entropy Audit: Uncertainty Should Be Measured And Actionable The entropy-audit wireframe expresses a core quality idea: knowledge artifacts have uncertainty, missing citations, verification gaps, and error-correcting work to do. Strategic implications: - Aquin should surface uncertainty instead of hiding it behind confident prose. - Missing citations, unsupported claims, stale evidence, and weak links should become actionable errors. - The system should recommend high-information-gain actions such as literature search, source verification, cross-source linking, or note repair. - Entropy and distortion metrics should be useful heuristics, not fake precision. - Export readiness should include information-integrity checks. ### 7. Sum-Over-Histories: Serious Questions Need Multiple Reasoning Paths The sum-over-histories wireframe frames inquiry as several possible reasoning paths with different confidence weights, followed by human selection. Strategic implications: - Aquin should support branching hypotheses rather than forcing premature convergence. - Agents can propose possible paths, but humans should collapse the working path when stakes require judgment. - Alternative paths should remain recoverable for later review. - Confidence should attach to paths, not just final answers. - The user should see why a path was selected, not merely the selected output. ### 8. Trail Compression: Preserve Full Fidelity, Publish Usable Insight The trail-compression wireframe shows a raw associative trail compressed into executive nodes, with provenance and human approval. Strategic implications: - Aquin should preserve the full-fidelity trail while also helping users produce compressed, useful summaries. - Compression should be explicit, reviewable, and reversible where possible. - Executive summaries should expose their source trail and compression report. - Agents can compress, but publication requires human approval when the compressed output will guide others. - Aquin should distinguish between raw reasoning traces, working notes, and shareable insight trails. ### 9. Aquinas Disputatio: Truth-Seeking Requires Structured Objection And Response The disputatio wireframe gives Aquin a classical structure for disciplined disagreement: objections, contrary authority, response, and replies to objections. Strategic implications: - Aquin should support formal argument workflows, not only freeform notes. - Good knowledge artifacts should survive objections. - Objections should be preserved in the artifact rather than deleted after resolution. - Responses should connect back to sources, principles, and evidence. - Human sign-off should be required for complete disputations. ## Operating Model: The Error-Correcting Knowledge Loop Aquin should organize serious knowledge work into a repeating loop: 1. **Choose the right problem** using Hamming-window prioritization. 2. **Form the inquiry** as a research thread with explicit purpose, scope, and stakes. 3. **Gather sources** through high-fidelity ingest with citations, hashes, and provenance. 4. **Map structure** with typed graph nodes and meaningful relationships. 5. **Branch hypotheses** through multiple reasoning paths. 6. **Challenge claims** through conflict agents, objections, counterexamples, and disputatio. 7. **Audit uncertainty** through entropy, distortion, missing-citation, staleness, and contradiction checks. 8. **Record judgment** through explicit gates, rationale, and sign-off. 9. **Compress insight** from full-fidelity trails into usable executive nodes. 10. **Export artifacts** with enough provenance, lineage, and review context to be trusted outside Aquin. 11. **Propagate learning** through lineage, supersession, replication, variation, and deprecation. This loop is the heart of the product. Every major feature should either strengthen the loop or be questioned. ## Core Principles 1. **Truth before fluency**. A polished answer is not enough; claims must be inspectable. 2. **Files are canonical**. OKF Markdown/YAML is the durable knowledge substrate. 3. **Projections are rebuildable**. Graphs, caches, mirrors, layouts, diagnostics, and summaries can be recreated from canonical files and project-local records. 4. **Provenance is non-negotiable**. Sources, transformations, review decisions, timestamps, and export evidence must remain inspectable. 5. **Human judgment remains in control**. Assisted changes do not silently become canonical. 6. **Agents propose; humans dispose**. Agentic workflows may suggest, diagnose, challenge, compress, or draft, but important changes require explicit acceptance. 7. **Uncertainty is a product surface**. Unknowns, contradictions, missing citations, weak links, and stale claims should be visible and actionable. 8. **Structured disagreement improves knowledge**. Objections, conflict, plausible error, and alternative paths are tools for truth-seeking. 9. **OKF serves humans and agents together**. One artifact should be readable, structured, parseable, and reusable. 10. **Local-first is a trust boundary**. Core work should function without cloud services. 11. **Local AI is a core capability**. A local model path is part of the product contract, not an optional afterthought. 12. **Quality beats volume**. Aquin should help users create clearer artifacts, not merely larger collections. 13. **Portability protects users and readers**. Knowledge should remain useful in standard tools. 14. **Time matters**. Knowledge has history, staleness, revision, lineage, and decay; timestamps and temporal views are first-class. 15. **Native quality is part of trust**. A serious knowledge tool must not feel like a temporary demo surface. 16. **Human flourishing is the north star**. Aquin should help people use knowledge to make better decisions, teach better, build better, and serve others better. ## Functional Requirements **FR-1: Native Local-First Foundation** Aquin must operate locally by default. Core authoring, validation, graph analysis, review, local AI assistance, and project work must not require external services. **FR-2: Swift-Native macOS Product Quality** Aquin should feel like a serious Mac document application: coherent save behavior, keyboard-first editing, native commands, accessibility, clear error recovery, and polished user-facing vocabulary. **FR-3: Canonical Portable Storage** Canonical knowledge lives in OKF Markdown/YAML files inside a folder-backed project package. Project-local databases, caches, indexes, and projections are supporting state. **FR-4: Human Authorship And Editing** Users must be able to write, revise, restructure, annotate, cite, and validate OKF files through a dedicated Mac editing experience while retaining ordinary file access. **FR-5: High-Fidelity Ingest** Aquin should ingest local and authorized external material into editable OKF derivatives or review proposals while preserving source identity, hashes, lineage, timestamps, profile information, and uncertainty. **FR-6: Provenance And Citation Discipline** Citations belong in the canonical OKF files when a note or claim needs them. Aquin's provenance mechanisms add machine-readable lineage, transformation history, and review evidence without silently rewriting human citations. **FR-7: Typed Graph And Projection Support** Graphs, search indexes, saved views, role filters, timelines, diagnostics, and quality checks are rebuildable projections over canonical OKF files and project-local workflow records. **FR-8: Review-Gated Assistance** Local AI, graph repair, internet-assisted drafts, ingest uncertainty, conflict resolution, compression, and export transformation must become proposals or reviewable drafts before they affect canonical knowledge. **FR-9: Conflict Agent And Error Injection** Aquin should support adversarial workflows that introduce plausible counterclaims, objections, missing assumptions, and failure cases so users can strengthen or revise their artifacts. **FR-10: Structured Disputation** Aquin should support an Aquinas-style disputatio workflow: objections, sed contra, corpus response, replies to objections, and human sign-off. **FR-11: Hamming Important Problem Window** Aquin should help users focus on work where importance to the field or humanity intersects with current tooling leverage. **FR-12: Rate-Distortion Research Budget Allocation** Aquin should help users allocate attention between agent execution and human judgment based on distortion, uncertainty, expected information gain, current energy, and the value of the work item. **FR-13: Entropy Audit And Error Correction** Aquin should surface entropy, missing citations, stale claims, weak evidence, contradictions, and unsupported links as actionable quality issues. **FR-14: Sum-Over-Histories Reasoning** Aquin should support multiple possible reasoning paths for a research question, preserve alternatives, attach confidence or uncertainty to paths, and require human judgment when selecting a working path. **FR-15: Trail Compression** Aquin should preserve full-fidelity associative trails while helping users compress them into executive trails, summaries, reports, and exportable insight nodes with provenance and human approval. **FR-16: Research Lineage And Cumulative Ledger** Research threads should record relationships such as extends, supersedes, replicates-with-variation, contradicts, deprecates, and supports. Status propagation should follow visible rules. **FR-17: Subset Selection And Refinement** Users must be able to select coherent subsets of an OKF project and refine them for a specific audience, purpose, capability, or downstream agent context. **FR-18: Exportable OKF Artifacts** Export must produce artifacts that remain useful outside Aquin. Export readiness should surface sovereignty, provenance, review, lineage, rebuildability, uncertainty, and portability signals before sharing. **FR-19: Local AI And Agentic Modes** Local AI should assist with metadata, graph diagnostics, quality checks, summaries, trail compression, conflict generation, draft proposals, and workflow coaching. More agentic workflows may exist, but they must remain scoped, provenance-aware, and human-reviewable. **FR-20: Explicit Network And Model Control** No project content should leave the user's machine unless the user explicitly authorizes that action for a specific workflow. External retrieval, external model calls, and sharing should record provenance and user intent. **FR-21: Temporal Evolution And ISO8601 History** Important events should carry clear timestamps, including proposal creation, acceptance, rejection, source ingest, judgment gates, export, supersession, deprecation, and status propagation. **FR-22: Agent-Readable Knowledge Contracts** OKF artifacts should eventually support durable context packs, declarative skills, routing hints, examples, domain terminology, playbooks, data-handling policies, evaluation cases, provenance, and change history for agent use. ## Functional Architecture ### Canonical Knowledge Substrate Aquin projects are folder-backed packages containing ordinary Markdown/YAML OKF files. The package may also contain project-local Aquin state under `.aquin/`, but the canonical knowledge remains visible and editable as files. ### Ingest Channel Ingest converts external sources into editable OKF derivatives or review proposals. The ingest path must preserve source identity, hashes, lineage, profile information, timestamps, and uncertainty. ### Human Editing Interface The editor is the first-class authoring surface. It should make Markdown editing, frontmatter guidance, validation, source linking, review state, and typed relationship editing easy without hiding the underlying file model. ### Review And Proposal Layer Assisted changes, graph repairs, internet drafts, ingest uncertainty, collaboration actions, conflict resolutions, and compression outputs should flow through durable proposal review. Acceptance mutates canonical Markdown only after explicit human action. ### Graph And Analysis Layer Graph views, saved lenses, search, diagnostics, quality checks, timelines, and role filters help users inspect structure, gaps, weak links, centrality, source relationships, and export readiness. These views must be rebuildable and explainable. ### Research Workflow Layer Research Planner and research-thread workflows help users manage inquiries, evidence dependencies, hypothesis branches, human judgments, disputations, audit findings, and artifact steps. These records should connect to documents, graph nodes, proposals, sources, and exports. ### Conflict And Disputation Layer Aquin should support conflict agents, provocative claims, objections, sed contra material, counterexamples, refutations, incorporation paths, and final replies. This layer exists to strengthen knowledge by exposing weaknesses before export. ### Entropy And Distortion Layer Aquin should monitor uncertainty, missing evidence, stale claims, review gaps, weak relationships, and expected information gain. Entropy and distortion should guide work selection, not pretend to be exact measures of truth. ### Attention Allocation Layer Aquin should help users decide when to spend human attention and when to delegate bounded work to agents. The product should bias human effort toward judgment, synthesis, creativity, prioritization, and values-sensitive choices. ### Trail Compression Layer Aquin should preserve raw associative trails and produce compressed executive trails only through visible transformation. The user should be able to inspect what was removed, what was preserved, and why the compressed trail is safe to share. ### Local AI Layer Local AI should be private by default and useful in bounded ways: metadata repair, graph diagnostics, provenance checks, summaries, contradiction checks, draft proposals, trail compression, and workflow coaching. Internet retrieval requires explicit user authorization and recorded source provenance. ### Export Layer Export is the product payoff. Aquin should help users package refined subsets with the evidence, metadata, uncertainty, lineage, and review context needed for another human or agent to trust and use them. ### Future Decentralized Provenance For public shared bundles, Aquin may eventually anchor content hashes, version links, signatures, timestamps, and change metadata on-chain or in other decentralized provenance systems while keeping full content off-chain in ordinary files or content-addressed storage. This is a future trust extension, not a requirement for local authoring. ## Product Doctrine For AI Aquin should use AI aggressively where it increases leverage and conservatively where it risks false authority. AI is appropriate for: - Metadata suggestions. - Draft summaries. - Graph diagnostics. - Missing-citation detection. - Contradiction and objection generation. - Plausible error injection. - Source comparison. - Trail compression drafts. - Export readiness checks. - Workflow coaching. AI must be gated for: - Canonical file mutation. - Claim creation that will be exported as knowledge. - Source interpretation where uncertainty is material. - Final research-thread status changes. - Disputation completion. - Compression publication. - External sharing. - External model calls. The product should make one distinction relentlessly clear: agent acceleration is not agent authority. ## Success Measures Aquin is succeeding when: - A user can open the project folder and understand the canonical knowledge without Aquin. - Another person can inspect an exported artifact and see sources, claims, review status, lineage, and uncertainty. - Agents can use OKF artifacts without stripping away human-readable meaning. - Important claims are cited, or explicitly marked as unsupported, contested, provisional, or judgment-based. - The system surfaces missing citations, stale claims, contradictions, and weak links before export. - Users spend more attention on high-importance, high-leverage work and less attention on mechanical cleanup. - Conflict workflows improve the artifact instead of merely producing debate. - Compressed summaries remain traceable to full-fidelity trails. - Research threads can supersede, extend, replicate, or deprecate prior threads without losing history. - Local-first trust is real, not just branding. ## Product Boundaries Aquin should avoid becoming: - A cloud-first collaboration suite. - A generic chat wrapper. - A decorative graph visualization toy. - A citation manager with AI bolted on. - A task manager detached from truth-seeking. - A surveillance workspace for monitoring people. - A system where agent output silently becomes knowledge. - A product that rewards volume over clarity. - A product that hides uncertainty to feel smarter. The product should be ambitious, but the ambition should remain disciplined: help people produce trustworthy knowledge artifacts that improve judgment and can be reused by others. ## Pedagogy And Tutorial-Driven Validation Aquin should teach better knowledge practice while users produce real artifacts. Tutorials are part of the product experience, but they do not replace the engineering roadmap. The tutorial strategy has three tracks: - Vision docs explain why Aquin exists and what principles it must satisfy. - Engineering roadmap docs define active implementation work under `docs/dev/active/`. - Tutorial and validation scripts teach workflows and classify findings as `Pass`, `Friction`, `Gap`, or `Vision risk`. Vision risks take priority because they threaten file sovereignty, provenance, explicit human review, rebuildable projections, local AI privacy, truthfulness, temporal integrity, or portable export. Gaps and friction become roadmap candidates only when they block or materially weaken the user's ability to produce high-quality OKF artifacts. ## Current Product Direction Near-term work should focus on: - Making `.okfproject` package behavior and export behavior coherent. - Removing stale archive and prototype-era concepts from active product paths. - Finishing native Mac polish, save semantics, accessibility, and product vocabulary. - Strengthening provenance, review, temporal history, and export readiness. - Making local AI a real product path for safe, bounded assistance. - Building graph projections that make role, lineage, confidence, and provenance inspectable. - Implementing explicit judgment gates for risky or meaning-changing workflows. - Adding conflict-agent and disputatio workflows after the canonical review layer is solid. - Continuing model extraction so the app architecture remains maintainable. - Running tutorial-driven validation and promoting only triaged findings into active roadmap work. ## Recommended Strategic Additions ### 1. Make Aquin An Error-Correcting Knowledge Workbench The strongest unifying idea across the wireframes is error correction. Aquin should not merely help users collect knowledge. It should help them find and repair epistemic defects: missing evidence, weak provenance, unsupported claims, stale beliefs, hidden contradictions, bad compression, and premature conclusions. ### 2. Treat Human Attention As A Sacred Resource The rate-distortion and Hamming-window concepts should become central to the product. Aquin should help users spend attention where it matters: high-importance, high-leverage problems where human judgment changes the outcome. ### 3. Make Disagreement A First-Class Workflow The conflict-agent and disputatio concepts should be product-defining. Aquin should make it normal to ask: What is the best objection? What would make this false? What evidence discriminates between these paths? What remains unresolved? ### 4. Build Trust Passports For Exported Artifacts Every serious export should eventually carry a trust passport: source summary, citation coverage, unresolved claims, review gates passed, compression history, lineage, timestamps, known limitations, and intended audience. ### 5. Give Agents A Durable Knowledge Plane OKF should become the durable knowledge plane for agents: not a prompt dump, but a portable, versioned, provenance-rich context package with human-readable meaning and machine-readable structure. ### 6. Make Temporal Evolution Visible Aquin should treat knowledge as historical. Claims are created, revised, challenged, strengthened, weakened, superseded, deprecated, and exported. Time should be visible in the interface and in the canonical records. ### 7. Avoid Fake Precision Entropy, distortion, confidence, and leverage are useful only if the product is honest about their meaning. These metrics should guide attention and review; they should not pretend to certify truth mechanically. ## Open Questions - What should the durable portable export shape be now that ZIP-backed `.okf` archives are no longer part of the core path? - Which review facts belong in portable `log.md` versus project-local review SQLite? - How far should local AI go before the workflow requires a stronger explicit human approval boundary? - Which graph diagnostics should become automatic quality gates before export? - How should Aquin quantify entropy, distortion, confidence, and information gain without creating fake precision? - Which typed edge verbs should be canonical at the beginning, and which should emerge from use? - How should research-thread status propagation work when a thread supersedes, extends, or replicates another thread? - What is the minimal viable trust passport for exported OKF artifacts? - When, if ever, should decentralized provenance move from technical extension to user-facing workflow? - How should OKF artifacts interoperate with live agent protocols without turning durable knowledge into volatile session state? ## Relationship To Prior Vision Work Older strategy, architecture, and diagram documents remain useful as historical references only when active docs explicitly promote their ideas. The current source of truth is this vision, the technical architecture and gap analysis under `docs/technical/`, and the active roadmap under `docs/dev/active/`. This is a living document. Update it when product decisions deepen, when implementation reality changes the architecture, or when validation shows that a principle needs clearer operational meaning.