flowchart LR
A["New source or gap"] --> B["Inventory record"]
B --> C["Access and storage policy"]
C --> D["Source note or backlog state"]
D --> E{"Existing boundary?"}
E -- "yes" --> F["Update chapter source queue"]
E -- "no" --> G["Propose precise chapter"]
E -- "defer" --> H["External literature queue"]
F --> I["Claim/evidence mapping"]
G --> I
H --> J["Backlog item with non-claims"]
I --> K["Proof/test/backlog item"]
J --> K
87 Open Research Agenda and Bibliography Plan
87.1 Chapter status
| Field | Value |
|---|---|
| Chapter ID | open-research-agenda-and-bibliography-plan |
| Part | Part IV - Evidence, Implementation, and the Living Book |
| Status | conceptual |
| Manuscript maturity | v0.3 semantically audited manuscript |
| Last updated | 2026-08-08 |
| Primary source records | 31 assigned records with 29 exact mappings, including relational_dimension_compiler and its completed 2026-07-25 chapter-boundary adjudication |
| Claim label | Design rationale |
| Evidence level | argument |
| Source queue | primary: verification_bandwidth, benchmaxxing; supporting: alignment_field, scf, planforge, vcm_public, spinoza, talos, rmi, cgs, genesiscode, simulation_scaling, field_of_god_ai_constitution, project_theseus_whitepaper, circle_calculus_core, circle_ai_contract_suite, theseus_circle_transfer, relational_dimension_compiler; connector/recovery: moecot, coilmoecot, road_to_agi, vcm_editable |
| Source loading state | source notes: verification_bandwidth, deterministic_capability_compilation, platonic_world_model, benchmaxxing, alignment_field, scf, planforge, vcm_public, spinoza, talos, rmi, cgs, genesiscode, simulation_scaling, moecot, coilmoecot, road_to_agi, vcm_editable, field_of_god_ai_constitution, project_theseus_whitepaper, circle_calculus_core, circle_ai_contract_suite, theseus_circle_transfer, ext_graphrag_2024, ext_hipporag_2024, ext_mmlu_2020, ext_gpqa_2023, ext_livebench_2024, ext_lean4_theorem_proving, ext_moe_llm_survey_2024, relational_dimension_compiler; raw cache: verification_bandwidth, benchmaxxing, alignment_field, scf, planforge, vcm_public, spinoza, talos, rmi, cgs, genesiscode, simulation_scaling, coilmoecot, road_to_agi; connector/recovery: moecot, vcm_editable |
| Test state | Source inventory validation, source-note validation, research_backlog_record.valid.json, and new_paper_triage_scenario.valid.json protocol fixture validation pass; the backlog fixture carries triage state, source storage policy, public-safety state, chapter action policy, chapter-decision refs, deduplication state, merge/split policy, boundary rationale, required pre-drafting work, evidence-transition preconditions, source refs, promotion blockers, support-state effect, and non-claims; the triage scenario fixture covers synthetic update-existing, propose-new-chapter, defer-external-literature, and reject-duplicate cases only; AsiStackProofs.BibliographyPlan retains the two derived negative cases that reject missing source records and accepted assignments to nonexistent chapters. |
87.2 Drafting guardrail
Future work needs its own managed layer. Unread sources are not cited, bibliography metadata is not inferred, and missing or private artifacts are not promoted as evidence.
Living-book methodology defines how the manuscript changes without losing evidence control. The final agenda disciplines that change after publication: every missing source, external-literature queue, proof target, experiment idea, artifact reproduction task, and possible chapter insertion becomes a backlog item with an owner boundary instead of a vague promise that future work will fill the gap.
The agenda is also admission control for future enthusiasm. A new paper, benchmark result, tool, proof idea, or conversation-mined insight can improve the book, but it should not enter as undifferentiated confidence. It enters as a source record, source note, chapter assignment, claim mapping, proof or test route, and residual.
87.3 Human Reading Path
Concrete lens. A static bibliography keeps every title and encourages inference from source count. The admission gate instead records the owner, direct-review state, deduplication result, blockers, and exact route before prose changes.
The research agenda prevents the architecture from hiding its unfinished work after the argument reaches its end. It names what remains missing: sources to load, literature to normalize, experiments to run, proofs to refine, schemas to harden, and claims that must not rise above their current support state. A serious architecture needs that backlog to be public enough that future work cannot quietly skip it.
This is not an apology for incompleteness. It is the living-book contract in its most explicit form. The architecture can keep improving only if open work stays visible instead of being smoothed away by confident prose. The final state should make uncertainty actionable: fetch, read, test, formalize, reproduce, revise, or close.
That turns the bibliography from a decorative ending into a control surface for the next revision. Sources are not trophies, and gaps are not embarrassments. They are routes for future work, each with a boundary, owner, evidence condition, and reason why the claim is not stronger yet. That routing discipline is how a living book stays corrigible over time.
87.4 Problem
A living book can decay into a pile of sources unless it has a research backlog. Future writing agents need to know which source to load, which chapter owns the boundary, which evidence is missing, which proof target is operational, and when a new paper should update an existing chapter instead of creating a duplicate one.
They also need to know what not to do. Do not cite a title because it sounds relevant. Do not add a chapter because a source uses new language for an old boundary. Do not promote a claim because a private artifact exists somewhere else. Do not turn an external reference named inside a draft into a citation until the reference itself has been checked. The research agenda is therefore a control surface, not a bibliography afterthought. It turns future curiosity into accountable intake decisions, source notes, merge rules, proof or test backlogs, public-safety choices, and explicit non-claims before new prose changes the architecture.
87.5 Why existing approaches are insufficient
A pile of sources or ad hoc citations does not tell future writing agents what to load, compare, prove, test, or defer. It also creates citation drift: a chapter may cite a source it has not actually mined, or a new source may be treated as evidence before it has an inventory record and source note.
The bibliography for this book is not just a list of references at the end. It is an operating surface for a living research program. Each source has to answer practical questions: which chapter owns it, which claim can it inform, what text has actually been read, what remains private or connector-only, what external literature needs direct verification, and what would have to happen before the source could raise support above argument. A static bibliography cannot carry that burden by itself.
The external backlog itself has baselines. MMLU (ext_mmlu_2020), GPQA (ext_gpqa_2023), and LiveBench (ext_livebench_2024) mark evaluation frontiers that need careful use; Lean (ext_lean4_theorem_proving) marks the proof-assistant lane; and the MoE survey (ext_moe_llm_survey_2024) marks routing literature that cannot be replaced by local terminology. The agenda records these as load queues and comparison surfaces, not as already-synthesized evidence.
The research agenda is the matching map of ignorance. It names the work the book is not yet allowed to pretend is done: citation normalization, benchmark reproduction, semantic proof adequacy, chapter-level tests, external literature review, artifact import, reader-edition review, and new-paper triage. Keeping that ignorance explicit is what lets the book keep growing without becoming more confident merely because it became longer.
The book therefore needs more than bibliography management. It needs state management for knowledge work: unread, inventoried, cached, source-noted, mapped to claims, passage-reviewed, integrated into chapter prose, tested, mechanized, reproduced, deprecated, or refuted. Those states are not decoration. They are the difference between a living book and a confident scrapbook.
87.6 Core Claim
Reader claim. The research agenda is the book’s admission desk: it decides whether new material updates an owner, creates a genuinely new boundary, waits for direct inspection, or is rejected as a duplicate.
Operational rule. No source changes prose or support until its identity, access and publication boundary, direct review state, chapter owner, claim mapping, deduplication decision, required work, blockers, and closure condition are recorded.
[open-research-agenda-and-bibliography-plan.core, label: Design rationale, support: argument] Open Research Agenda and Bibliography Plan owns a research-program-, source-or-gap-, backlog-item-, access-, provenance-, public-safety-, chapter-boundary-, claim-, proof-or-experiment-, deduplication-, evidence-transition-, owner-, next-action-, closure-, consumer-, environment-, and time-specific Research Backlog Admission and Closure Contract. Every new paper, local project, missing artifact, conflicting result, proof idea, experiment, reproduction need, correction, or chapter proposal enters through exact intake, triage, assignment, preconditions, blockers, non-claims, and terminal closure before it changes prose or support. No title, citation, source count, inventory row, source note, queue, priority label, backlog size, fixture, theorem, validator, or completed reading task alone establishes citation accuracy, literature completeness, research quality, claim support, reproduction, transfer, AGI, ASI, or SOTA.
The claim remains at argument support. The repository now has source inventory, source notes, chapter listings, exact core-claim source-note mappings, and passage-reviewed mapping refs for the assigned sources, but external literature normalization, direct citation checks, artifact reproduction, and new-paper triage rehearsals remain incomplete.
87.6.1 Claim-source mapping status
Appendix C now records twenty-nine exact passage-reviewed mappings across thirty-one assigned sources for this claim. Seven mappings close prior prose-only gaps for GraphRAG, HippoRAG, MMLU, GPQA, LiveBench, Lean 4, and the MoE survey; the Relational Dimension Compiler mapping now records an admitted argument-level chapter and an unexecuted research program. The review turns the agenda into a context-loading map for future proof and experiment work; it does not turn chapter admission into completed evidence.
| Source group | Reviewed support | Boundary |
|---|---|---|
Verification, evidence, and benchmark backlog: verification_bandwidth, benchmaxxing, spinoza |
Context-adequacy tests, semantic-unit verification, contradiction-rate proposals, benchmark lifecycle, residuals, ledgers, anti-Goodhart safeguards, proof-carrying claims, belief revision, downgrade paths, and protected axioms. | No contradiction-rate test, theorem result, benchmark run, mutation/holdout run, contamination audit, verifier-quality result, autoformalization result, or citation-normalized literature claim. |
Governance, replacement, and safety backlog: alignment_field, scf, field_of_god_ai_constitution |
Normative assumptions, value conflict, dignity, corrigibility, stable capability identity, evaluator boundaries, authority ceilings, lifecycle ordering, rollback, recovery, constitutional tool/memory/self-improvement constraints, and least sufficient power. | No empirical consciousness, moral-status, alignment, rights-preservation, route/evaluator correctness, replacement safety, runtime policy-engine, red-team, system-prompt, or moral-correctness result. |
Planning, context, execution, and improvement backlog: planforge, vcm_public, talos, rmi, genesiscode |
Planning experiments, primitive schemas, VCM packet schemas, source binding, typed job lifecycles, audit/replay, residual escrow, specialist lifecycle, capability policies, effect logs, replay, and translation-validation topics. | No planner implementation, VCM-Bench, retrieval-quality result, Talos runtime/security run, RMI prototype, GenesisCode implementation, replay checker, benchmark, or security audit. |
Compression, simulation, and optional substrate backlog: cgs, simulation_scaling, coilmoecot, vcm_editable |
Compact-seed and residual-honesty research items, simulation contracts and resource limits, optional routed coil/specialist lanes, richer VCM terminology, planner-guided paging, and third-party citation candidates. | No CGS benchmark, simulation benchmark, feasibility audit, CoilMoECOT baseline, route run, VCM model-facing result, or normalized third-party citation. |
Implementation and proof-contract backlog: moecot, road_to_agi, project_theseus_whitepaper, circle_calculus_core, circle_ai_contract_suite, theseus_circle_transfer |
Runtime artifact tracking, readiness gates, replay, promotion blockers, remaining-work status, report-first Theseus artifacts, theorem ids, proof-boundary discipline, AI contract fields, receipts, consumer gates, and transfer-lane insertion rules. | The separate Circle rope receipt slice covers one bounded external receipt replay only; no MoECOT runtime artifact, reproduced MoECOT command, current Theseus report verification, broad Circle contract pack, ASI Stack consumer gate, transfer consumer, proxy benchmark, private workload report, or model-improvement result exists. |
87.6.2 Contribution-owned derivative outlines
Derivative publication work is admitted only when it has a contribution owner, an evidence boundary, and a reason to exist outside the full book. The current program has three such owners:
| Derivative | Central question | Required manuscript components | Evidence gate before release |
|---|---|---|---|
| Governed-cognition interface contracts | Can explicit noninheritance and typed transitions improve controllability without destroying useful throughput? | Problem and comparator taxonomy; plain, normative, and machine projections; transition calculus; reference trace; strongest simpler alternatives; trust and liveness costs; bounded conclusions. | Semantic P0–P6 audit for consumed formal artifacts plus a terminal matched natural repository-change result. |
| Public claim-state transition discipline | Can a public technical work change its claims without laundering source, proof, experiment, or release state? | Claim identity; support-state transitions; defeaters and downgrades; correction propagation; product/release boundaries; failure cases; maintenance cost. | Natural substantive change stream with observed corrections and no unsupported promotion, plus exact public-state reconciliation. |
| Record/reality reconciliation and residual honesty | Which observations are needed before records about plans, execution, rollback, or release may be treated as external truth? | Plan/effect and receipt/reality separations; observation contracts; effect-complete rollback; residual conservation; public-surface incidents; recovery and expiry. | Named implementation and independent observation on reachable effects, including partial rollback and stale-public-surface challenges. |
The governed repository-change flagship receives a methods-and-results paper only after its outcome is terminal. Negative, mixed, or inconclusive results remain publishable when the instrument is competent and the denominator, attacks, costs, and failed gates are intact. Until those gates pass, the outlines are an organizational device for the book’s final conclusions, not submission-ready papers or evidence of novelty.
87.6.3 Relational Dimension Compiler research program
relational_dimension_compiler is passage-reviewed and mapped across its nine adjacent owner interfaces. It also clears the conceptual distinct-interface test for relational-dimension-compilation-and-polyadic-cognition chapter: dimensional typing, semantic/computational/storage arity, persistent role-sensitive relation objects, adaptive relational-order routing, relation qualification lifecycle, reversible semantic contraction, and RODIE form one coherent artifact and failure boundary not owned elsewhere.
The 2026-07-25 full-coverage audit found that the former structural freeze was no longer active and that continued deferral contradicted the book’s no-deferral policy. The chapter is now admitted after Replaceable Cognitive Substrates with a complete argument-level manuscript, manifest contract, mapped evidence sources, falsification program, and explicit non-claims. Admission changes ownership, not evidence: no RDC, RODIE corpus, natural-task advantage, independent evaluator, reproduction, or support promotion exists.
The source packet creates five coordinated research lanes:
- verify the paper’s external bibliography source by source rather than inheriting its citations through the Corben-authored paper;
- formalize finite typed role reification at the narrow ceiling of representational reconstruction;
- specify relation request, packet, axis, incidence, order-decision, and contraction-certificate records with adversarial mutations;
- build an anti-self-confirming RODIE pilot with strong lower-order rescues, proposal denominators, held-out roles/topologies/branches/scales, and total cost; and
- transfer any surviving mechanism to one natural workload and reproduce it independently before broadening the claim.
Existing chapters receive only the pieces they already own. The governed backlog and triage records preserve the prospective owner so these bounded integrations do not erase the paper’s distinctive conjunction or create nine competing RDC chapters.
87.7 Mechanism
87.7.1 Reference-use trace: four ways a source can enter without becoming evidence
This chapter is a reference owner, so its concrete unit is an intake decision rather than an invented anecdote. The public fixture new-paper-triage-demo-001 sends four synthetic cases through the same gate:
| Intake condition | Decision | Why |
|---|---|---|
| A source refines context-adequacy vocabulary but introduces no new owner, invariant, or failure boundary. | Update the existing chapter. | Create the source record and exact claim mapping; do not multiply chapters for new terminology. |
| A source appears to own a distinct interface, invariant, artifact, and failure mode. | Propose a new manifest chapter. | First prove that no current owner can absorb the boundary and define its source and test queues. |
| A secondary source names an external paper that has not been retrieved. | Defer. | A title is not support. Retrieve and directly inspect the work before citing or drafting from it. |
| A new export repeats an inventoried source without new claims or artifacts. | Reject the duplicate. | Preserve only version differences; do not duplicate evidence or prose. |
All four outcomes leave support unchanged. That is the point of the reference chapter: it turns “we found something interesting” into a reproducible routing decision. The simpler bibliography baseline stores every title and lets later writers infer importance from source count. This gate instead records why the item may change one existing chapter, justify a new boundary, remain unread, or close without action.
The fixture validates decision shape only; it does not claim these hypothetical sources were read or that the four decisions are universally correct. A real intake closes only after the named source, mapping, drafting, validation, and evidence-transition work has been performed.
The owned object is a Research Backlog Admission and Closure Contract, not a source list. Its lifecycle is:
| Phase | Required operations |
|---|---|
| 1. Freeze and identify | Freeze program purpose, intake window, source and gap classes, chapter graph, authority, access, rights, public safety, consumers, environment, time, and non-claims; identify every source, version, passage, citation, project, gap, claim, decision, proof, experiment, reproduction, owner, action, and closure. |
| 2. Inventory and classify | Record provenance, metadata source, acquisition, access, storage, checksum or version, permission, public safety, and canonical note; keep unread, cached, noted, mapped, reviewed, integrated, tested, mechanized, reproduced, inaccessible, private, rejected, duplicate, superseded, deprecated, refuted, and retired states distinct. |
| 3. Separate and extract | Separate Corben/local, conversation-intent, external, benchmark, proof, private, and missing-source classes; extract thesis, mechanisms, evidence, failures, chapters, claims, contrary evidence, citation data, and limits before drafting or support use. |
| 4. Triage and bind | Route to update, propose a distinct owner, appendix, proof or experiment work, defer, reject, or retire; apply the distinct-interface and merge/split tests; bind assignments to passages, atoms, rationale, alternatives, non-claims, dependencies, and reviewer authority. |
| 5. Execute and verify | Create lane-specific proof, executable, empirical, causal, reproduction, transfer, and normative work; record pre-drafting and transition conditions; deduplicate with lineage; verify citation metadata, passage support, corrections, retractions, rights, and availability; rehearse natural conflict and failure cases. |
| 6. Close and maintain | Close only on consumed action, artifact, review, citation or chapter decision, transition or no-change record, residual, cost, and non-claims; propagate corrections and failed reproduction; monitor coverage, errors, duplicates, age, closures, burden, opportunity cost, and total cost; expire changed contracts. |
Admission is deliberately stricter than collection. A source or idea enters with an identity, access state, rights boundary, intended consumer, and declared reading depth. Classification says whether it is merely inventoried, actually read, passage-reviewed, mapped, integrated, tested, reproduced, contradicted, or retired. Extraction records both the strongest useful contribution and the conditions it cannot establish. Only then may triage update an existing owner, propose a distinct boundary, create proof or experiment work, defer the item, or reject it with preserved lineage.
Closure is likewise a consumer event rather than a checked box. A mapping is not closed until its claim owner acknowledges it; an experiment is not closed until artifacts, failures, denominators, and interpretation are adjudicated; a citation is not closed while identity or correction state is unknown. Every terminal state retains cost, residuals, affected consumers, and reopening conditions. This keeps an inaccessible source, failed reproduction, duplicate record, or negative result useful without allowing backlog volume or closure rate to masquerade as research quality.
Any item can stop honestly at inaccessible, private, deferred, duplicate, rejected, failed, refuted, or retired. Keeping that state is more valuable than converting the item into a confident paragraph. New vocabulary normally updates an existing owner; only a genuinely distinct interface, invariant, artifact, authority, or failure boundary can justify a chapter proposal.
What the research agenda route shows: The research agenda routes new material through inventory, access policy, source-note or backlog state, chapter-boundary decisions, claim mapping, and proof/test backlog. A deferred source remains external literature or a backlog item with non-claims instead of becoming unsupported chapter evidence.
A source backlog record tracks source id, access state, source storage policy, public-safety state, assigned chapters, chapter-decision refs, deduplication state, merge/split policy, source-note state, claim-mapping state, passage-review state, external-literature need, proof/test backlog, pre-drafting requirements, insertion decision, promotion blockers, and residuals.
For large cross-layer papers, “read” is not a sufficient terminal state. The July 2026 mining audit adds a section-family coverage ledger: every substantive section must resolve to chapter prose, a source-note synthesis, an experiment or roadmap obligation, or an explicit non-goal/defer decision. Unmapped sections, unsupported evidence inheritance, and paper-shaped duplicate chapters block closure. deterministic_capability_compilation and platonic_world_model are the first real intakes held to that stronger standard.
87.8 Interfaces
The contract joins twelve interfaces: human research and closure authority; inventory, caches, notes, connectors, projects, permissions, and acquisition; separate Corben/local, external, and author-intent appendices; manifest, outline, chapters, queues, citations, bibliography, and changelog; claim reviews, Appendix C, transitions, contrary evidence, residuals, and non-claims; proof, Lean, schemas, tests, benchmarks, studies, and reproduction artifacts; artifact, supply-chain, data, security, rights, retention, deletion, and replay controls; roadmap, living-book, stewardship, work-contract, status, terminal, and successor continuity; internal and local project lineage; external challenge queues; release, citation, reader, archive, site, and correction surfaces; and researchers, readers, AI agents, reviewers, citers, implementers, evaluators, downstream repositories, institutions, and future editions.
Each join preserves its own identifier, authority, version, acknowledgement, error, and expiry rather than treating a nearby repository path as successful delivery. A consumer that cannot resolve or accept the record remains an open backlog edge with an owner and next action.
87.9 Invariants
The eighteen invariants require exact scope; no unread or inaccessible source support; authoritative or unknown metadata; distinct provenance classes; distinct lifecycle states; passage-bound material claims; existing or distinct owner assignments; lineage-preserving deduplication; accepted evidence transitions only; separate work lanes; public-safety and privacy; retained negative and rejected items; consumed closure artifacts and decisions; correction propagation; versioned citation state; joint usefulness, failure, burden, and cost; exact synthesis and transfer scope; and a strict boundary around what inventory, note, mapping, fixture, theorem, validator, citation, and backlog counts can prove.
An expired, corrected, retracted, superseded, or consumer-rejected item cannot remain silently active. Closure conserves unresolved obligations, affected consumers, and residual ownership, while later reopening preserves the original decision and lineage instead of rewriting research history.
87.10 Failure modes
The named falsifiers are citation drift; bibliographic hallucination; unread-source laundering; inventory theater; anthology drift; duplicate-evidence inflation; source-result inheritance; private-evidence leakage; indirect-reference laundering; triage theater; backlog rot; activity laundering; negative and conflict suppression; closure laundering; correction failure; priority capture; research overhead and false deferral; and unsupported transfer. Each failure leaves a blocked citation, rejected assignment, duplicate link, correction, retraction, residual, no-change decision, failed reproduction, retired item, or reopened consumer. Bibliographic presence is never evidentiary force.
The contract also fails when its own burden hides more useful research, when privileged access or outcome-aware selection makes comparisons unfair, or when a clean record masks false metadata, weak reading, invalid proof, defective experiments, or unacknowledged downstream use. A high closure rate can therefore be evidence of premature rejection or closure laundering rather than program quality.
87.11 Minimum Viable Implementation
The implemented minimum contains twenty-eight exact source mappings; a source inventory and source-note corpus split across Corben/local, external, and author-intent appendices; one research-backlog record schema and public-safe fixture; and one new-paper triage scenario covering update-existing, propose-new-owner, defer-unread-external, and reject-duplicate decisions. Source inventory, note, protocol, appendix, citation, and mapping validators enforce the bounded record surface.
Two proof targets are implemented by two derived Lean counterexample declarations over hand-authored finite source-ingestion and chapter-assignment records. They reject a source-derived claim record with no ingested artifact and an accepted assignment to a nonexistent chapter. The direct projection restatements of those predicates were retired because they merely repeated their premises. The active backlog also retains citation normalization, artifact reproduction, proof and experiment work, GraphRAG/HippoRAG overlap, inaccessible sources, and rejected or blocked items.
No external-literature completeness, citation-accuracy study, natural intake campaign, real backlog-closure study, research-quality result, independent reproduction, transfer, or chapter-core support effect exists.
87.12 Mature Research Target
The mature argument-exit campaign must prospectively freeze a long-running natural research stream of new papers, local projects, inaccessible and private sources, conflicting findings, corrections and retractions, benchmark and proof artifacts, reproduction attempts, negative results, duplicate claims, and chapter proposals handled by multiple independent human and AI researchers. It must compare the contract with matched reference-manager, systematic-review, evidence-map, issue-tracker, wiki, ordinary bibliography, ad hoc agent-research, and strongest conventional research-program workflows under equal sources, access, tools, time, and review opportunity.
The campaign must inject metadata errors, title-only sources, duplicate variants, hidden conflicts, private leakage pressure, stale citations, retractions, boundary ambiguity, chapter novelty pressure, failed reproduction, correction propagation, and backlog rot. It must jointly measure citation precision and recall, source coverage and calibration, decision and claim accuracy, duplicate detection, error and conflict discovery, time-to-usable-evidence, closure quality, false admission and deferral, correction latency, reproduction yield, chapter churn, reviewer burden, compute, storage, opportunity cost, residuals, and total cost. Causal ablations must identify useful and harmful fields or gates.
Independent researchers, reviewers, repositories, institutions, disciplines, legal regimes, languages, and time periods must reproduce and transfer positive, negative, null, narrowed, blocked, and refuted results. Until then the contract remains a target architecture, not proof of literature completeness, citation accuracy, research quality, superior prioritization, claim truth, reproduction, transfer, or SOTA research management.
87.13 Codex test plan
| Test | Purpose | Status |
|---|---|---|
| Research backlog record fixture validation | Validate that a backlog record names source/gap, triage state, access state, source storage policy, public-safety state, assigned chapters, chapter action policy, chapter-decision refs, deduplication state, merge/split policy, boundary rationale, source-note state, claim mapping state, external-literature need, proof/test backlog, required pre-drafting work, evidence-transition preconditions, insertion decision, source refs, promotion blockers, support-state effect, residuals, next action, and non-claims. | implemented; passing via python3 scripts/validate_protocol_examples.py |
| Source inventory validation test | Check that assigned source IDs exist before chapter claims use them. | implemented by validation; validated locally |
| Source-note backlog audit | Check which assigned sources still lack usable source notes. | source notes, exact core-claim mappings, and passage-reviewed mapping refs currently exist for assigned sources; external-literature normalization and evidence transitions remain incomplete |
| Source-derived claim evidence gate | Check that a source-derived claim without a source note or ingested-source artifact is rejected at the finite record layer. | implemented by the retained derived counterexample source_derived_claim_without_source_record_rejected; citation-accuracy and source-interpretation quality not proved |
| New-source assignment gate | Check that an accepted new-source assignment cannot target a nonexistent chapter id. | implemented by the retained derived counterexample accepted_new_source_assignment_to_nonexistent_chapter_rejected; real new-paper triage quality not proved |
| New-paper triage scenario | Check whether a synthetic intake case updates an existing chapter, proposes a new boundary, defers unread external literature, or rejects a duplicate variant without changing support state. | implemented by protocol validation; validated locally |
87.13.1 Formalization hooks
| Tag | Module | Target | Status |
|---|---|---|---|
lean:bibliography.plan.operational_invariant |
AsiStackProofs.BibliographyPlan |
A source-derived claim with neither a source note nor an ingested artifact fails the finite source-evidence predicate. | implemented |
lean:bibliography.plan.failure_blocks_promotion |
AsiStackProofs.BibliographyPlan |
An accepted new-source assignment to a nonexistent chapter fails the finite assignment predicate. | implemented |
These Lean hooks are implemented as two derived negative-case theorems over finite source-derived-claim and new-source-assignment predicates. They prove only that the authored predicates reject two exact counterexamples: a source-derived claim with neither declared source-record surface, and an accepted assignment whose declared chapter does not exist. They do not prove citation accuracy, external-literature completeness, claim-to-source mapping quality, or new-paper triage quality.
87.14 Source crosswalk
The research agenda intentionally keeps a broad source set because it governs the backlog for the whole book. Appendix G remains the generated appendix for Corben’s corpus and local-project records. Appendix H remains the separate generated external-literature map. Appendix I remains author-intent and lineage context, not evidence.
Priority backlog:
- Claim-to-mechanism reconciliation beyond this core-claim mapping pass, especially where future chapter drafting depends on narrower mechanisms rather than broad source-family assignment.
- Live new-paper intake rehearsal against a user-provided source after source access, public-safety policy, and source-note permissions are clear.
- External literature normalization and direct citation checks.
- Reproduction or import of any benchmark artifacts before support promotion.
- Lean triage for proof targets that are currently schema or process contracts.
- New-paper triage rules for dynamic chapter insertion.
- Durable-semantic-memory overlap testing against
ext_graphrag_2024andext_hipporag_2024, including global sensemaking, associative navigation, provenance, contradiction, update, deletion, and poisoning boundaries before the provisional chapter can pass its add-or-reject gate.
Recovery boundaries:
- MoECOT, Road To AGI, VCM editable, and CoilMoECOT are readable through connector/source notes, but reported benchmark or runtime claims remain unpromoted.
- Private empirical results must stay out of public claims until artifact, command, environment, and publication permission are recorded.
- External references named inside source documents must be verified directly before becoming bibliography citations.
87.14.1 Manifest source assignment reconciliation
These rows keep Open Research Agenda and Bibliography Plan’s manifest assignments visible at their recorded review boundary. Passage review does not establish local reproduction, performance, safety, deployment, or support-state movement.
| Source | Intake role | Boundary |
|---|---|---|
verification_bandwidth |
Passage-reviewed Corben architecture source: Verification Bandwidth in Bounded Contexts. Supports a research backlog for context adequacy, semantic-unit verification, contradiction-rate tests, summary-loss accounting, and the distinction between long context and verified reasoning. | The proposed tests, theorem sketches, contradiction-rate claims, and adequacy claims remain source-proposed until mechanized or empirically tested. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
deterministic_capability_compilation |
Passage-reviewed Corben architecture source: Deterministic Capability Compilation: A Capability-Preserving Ladder from Executable Scaffolds to Governed Adaptive Agents. Corben-authored July 2026 architecture and research program for compiling executable scaffolds into contract-bound experts and linked Neural Capability Objects while retaining semantic obligation mass balance, candidate-specific translation validation, fallback, residual escrow, authority ceilings, reification, and effect-complete recovery. Existing chapters are upgraded first; no foundry implementation, learned-capability result, preservation result, safety result, SOTA result, AGI, ASI, or support-state promotion is inferred. | No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
platonic_world_model |
Metadata-first comparator: The Platonic World Model: A Semantic Constitution for Grounded, Proof-Carrying, Self-Editing Artificial Intelligence. Corben-authored July 2026 conceptual architecture and falsifiable research program for semantic continuity through stable Form lineages, immutable semantic versions, typed Essence Contracts, six mutually constraining planes, explicit proposition-attestation-commitment-proof separation, branch-protected world dynamics, qualified grounding, semantic transactions, runtime packet compilation, and federated mappings. Existing chapters are upgraded first; no implemented substrate, benchmark result, philosophical solution to grounding, safety result, SOTA result, AGI, ASI, or support-state promotion is inferred. | No passage-level source claim, local implementation, reproduction, safety, performance, deployment, support-state, or ASI result is established by this reconciliation row. |
benchmaxxing |
Passage-reviewed comparator: Benchmaxxing: The Performance Ratchet. Supports bibliography and agenda tracking for benchmark lifecycles, wall diagnosis, anti-Goodhart safeguards, benchmark ledgers, mutation, holdouts, contamination audits, and source-reported benchmark claims. | Does not provide local benchmark results, benchmark-harness validation, mutation/holdout runs, contamination audits, or citation-normalized external literature by itself. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
alignment_field |
Passage-reviewed Corben architecture source: Field of God / Alignment Field family. Supports tracking normative and metaphysical assumptions, rights, value-conflict boundaries, dignity, corrigibility, and needed external literature before public consciousness or moral-status claims. | Not empirical evidence for consciousness, alignment, moral status, rights preservation, dignity preservation, or value-conflict resolution. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
scf |
Passage-reviewed comparator: Stable Capability Fields. Supports research agenda items for stable capability identity, evaluator boundaries, qualification, route validation, lifecycle ordering, rollback, recovery, and authority non-escalation. | Does not prove SCF production safety, route/evaluator correctness, replacement safety, rollback success, recovery behavior, or authority-non-escalation runtime enforcement. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
planforge |
Passage-reviewed comparator: PlanForge. Supports future planning experiments around primitive schemas, dependency precedence, scheduler behavior, tier routing, fallback, replanning, and planner/executor boundaries. | No PlanForge implementation, scheduler run, route trace, planning benchmark, savings reproduction, learned decomposer, or execution handoff is validated here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
vcm_public |
Passage-reviewed comparator: Virtual_Context_Memory_v1. Supports research agenda items for VCM packet schemas, source binding, adequacy/admission separation, certificates, taint, revocation, invalidation, and context-system conformance tests. | Does not establish broad model-facing VCM gains, retrieval-quality improvement, context-system superiority, certificate truthfulness, or neighbor-system superiority. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
spinoza |
Passage-reviewed comparator: Proof of Belief / The Spinoza Architecture. Supports backlog work for proof-carrying claims, belief revision records, contradiction detection, downgrade paths, protected axioms, and formalization-scope boundaries. | Does not solve natural-language formalization, verifier quality, theorem validity for arbitrary claims, proof-object adequacy, or whole-system epistemic correctness. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
talos |
Passage-reviewed comparator: Talos Protocol. Supports research agenda items for typed job lifecycles, control planes, evidence records, Digital SCIFs, audit logs, replay, approval gates, runtime adapters, and delivery evidence. | No Talos runtime, benchmark, security enforcement, approval-service behavior, replay behavior, or Digital SCIF containment claim is reproduced here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
rmi |
Passage-reviewed comparator: Ratcheting Modular Intelligence. Supports backlog items for residual escrow, benchmark pressure, routed specialist attempts, arm/router lifecycle, regression preservation, quarantine, and modular improvement tests. | No RMI prototype, benchmark run, routed-specialist evaluation, router benchmark, regression-preservation result, or deployed modular-improvement loop has been independently verified here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
cgs |
Passage-reviewed comparator: Compact Generative Systems. Supports compression and residual-honesty research items for compact seeds, rule systems, residual channels, verification contracts, governance interfaces, hidden complexity debt, and CGS metrics. | No local CGS benchmark, utility test, proof of compact adequacy, generator, implementation, interpretability result, or hidden-residual absence result has been run. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
genesiscode |
Passage-reviewed comparator: GenesisCode. Supports agenda items for AI-assisted programming evidence boundaries: tiny kernels, semantic patches, provenance hashes, effect logs, capability policies, obligations, replay, and translation validation. | GenesisCode is not implemented, benchmarked, audited, replay-checked, security-audited, or mechanically verified in this repository. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
simulation_scaling |
Passage-reviewed comparator: The Simulation Scaling Law: Resource Constraints on Scope, Clockspeed, and Effective Fidelity in Nested Physical Simulations. Supports research agenda items for simulation contracts, scope, clockspeed, fidelity, efficiency, physical capacity, bottlenecks, and resource-accounted synthetic evaluation. | No physical experiment, simulator result, simulation benchmark, feasibility calculator, resource-bound sanity check, approximation audit, or independent literature audit was run here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
moecot |
Passage-reviewed comparator: MoECOT-Agent Architecture Whitepaper. Supports agenda tracking for MoECOT runtime artifacts, readiness gates, ledgers, replay, benchmark artifacts, residuals, promotion blockers, and implementation-reference gaps. | Reviewed source note only because authenticated connector text is not published here; reported MoECOT runtime artifacts, benchmark artifacts, ledgers, readiness gates, replay, and promotion claims remain uninspected and unreproduced. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
coilmoecot |
Passage-reviewed comparator: CoilMoECOT Whitepaper v2.0. Supports optional routed-substrate research agenda items for coil specialist lanes, readiness gates, benchmark baselines, residual handling, and cyclic/routed lane adoption. | Reviewed source note only because the local cache is an auth-gate placeholder; no CoilMoECOT route run, local benchmark, baseline comparison, readiness result, mutation/promotion result, or performance claim is validated. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
road_to_agi |
Passage-reviewed comparator: Road To AGI. Supports agenda tracking for MoECOT remaining work, reported benchmark commands/results, readiness status, promotion blockers, and implementation sequencing. | Reviewed source note only because the local cache is an auth-gate placeholder; roadmap/status context only, and reported commands, benchmark results, readiness state, and implementation status were not reproduced. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
vcm_editable |
Passage-reviewed comparator: Virtual_Context_Memory_v1.0_Editable. Supports bibliography triage for richer VCM terminology, planner-guided paging, evidence-carrying context, and external-literature candidates that need normalized metadata before citation. | Reviewed source note only because raw connector text is not published here; connector-readable variant only, and no VCM-Bench result, model-facing result, or citation-normalized third-party source is implied. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
field_of_god_ai_constitution |
Passage-reviewed comparator: Field of God AI Constitution. Supports agenda items for constitutional alignment clauses, tool-risk tiers, memory as power, least sufficient power, consent, audit, reversibility, red-team scenarios, and self-improvement freezes. | Constitutional specification only; no runtime policy engine, red-team suite, deployed governance behavior, Lean proof, system-prompt evaluation, or moral-correctness result has been run here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
project_theseus_whitepaper |
Passage-reviewed comparator: Project Theseus Whitepaper. Supports backlog tracking for public-safe Theseus reports, residual ledgers, benchmark ledgers, sparse teacher governance, trusted-node task allowlists, and report-first implementation artifacts. | Does not provide current Theseus report verification, command reruns, dashboard inspection, node state, model artifacts, reproduced benchmark results, or public-compute readiness evidence in this repo. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
circle_calculus_core |
Passage-reviewed comparator: Circle Calculus. Supports bibliography/proof-boundary agenda items for theorem ids, Lean proof status, theorem manifests, Python reference models, Rust utilities, Quarto proof surfaces, and finite cyclic contract receipts. | A separate external Circle rope receipt slice is recorded in docs/circle_external_receipt_slice.md, but no broad Circle theorem replay, sidecar regeneration, ASI Stack consumer gate, model-quality result, transfer result, or deployment result is implied. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
circle_ai_contract_suite |
Passage-reviewed comparator: Circle Calculus AI Contract Suite. Supports research agenda items for proof-carrying AI contract fields, receipt validation, theorem-linked recommendations, consumer gates, fingerprints, minimum evidence fields, and explicit non-claims. | No Circle contract pack, CLI command, Lean build, receipt replay, fingerprint check, downstream consumer, acceptance-policy check, or model-quality evidence was run here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
theseus_circle_transfer |
Passage-reviewed comparator: Theseus Circle Calculus Transfer Lane. Supports backlog and insertion rules for transfer-lane artifacts that separate structural smoke checks from quality, runtime, memory, transfer, and model-improvement claims. | No transfer consumer, structural smoke workload, proxy benchmark, private workload report, ordinary baseline, negative control, runtime measurement, memory measurement, transfer result, or model-improvement claim has been executed here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
ext_mmlu_2020 |
Passage-reviewed comparator: Measuring Massive Multitask Language Understanding. Provides an external benchmark-queue comparator for broad multitask knowledge evaluation, subject coverage, aggregate reporting, and limits on general-capability interpretation. | No MMLU model, run, score, contamination audit, baseline, reproduction, claim transition, or bibliography closure is established here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
ext_gpqa_2023 |
Passage-reviewed comparator: GPQA: A Graduate-Level Google-Proof Q&A Benchmark. Provides an external benchmark-queue comparator for expert-level, contamination-resistant science questions, human/model baselines, and evaluator difficulty. | No GPQA questions, model, run, score, contamination check, expert review, reproduction, support transition, or agenda closure is established here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
ext_livebench_2024 |
Passage-reviewed comparator: LiveBench: A Challenging, Contamination-Limited LLM Benchmark. Provides an external comparator for living benchmark refresh, objective scoring, contamination pressure, versioned evaluation, and historical result lineage. | No LiveBench data, model, run, score, refresh process, contamination test, reproduction, or research-backlog closure is performed here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
ext_lean4_theorem_proving |
Passage-reviewed comparator: Theorem Proving in Lean 4. Provides an external comparator for the proof-assistant backlog: theorem statements, definitions, tactics, kernel checking, reusable libraries, and the boundary between formalization and empirical truth. | Does not prove any ASI Stack theorem, validate the chosen models, close proof adequacy, establish runtime refinement, or complete a research item. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
ext_moe_llm_survey_2024 |
Passage-reviewed comparator: A Survey on Mixture of Experts in Large Language Models. Provides an external routing-literature comparator for sparse mixture-of-experts architectures, routers, expert specialization, load balancing, training and inference tradeoffs, and open problems. | No surveyed MoE system, router, model, benchmark, efficiency result, specialization result, reproduction, or local routing claim is established here. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
relational_dimension_compiler |
Passage-reviewed comparator: The Relational Dimension Compiler: Adaptive Polyadic Cognition with Bounded Computational Arity and Unbounded Semantic Structure. Adds a coordinated research program for typed polyadic cognition, relational-order routing, role-sensitive reification, object-field coupling, topology learning, semantic contraction, hardware compilation, schema evolution, and RODIE. | The Corben-authored bibliography and proposed experiments require source-by-source external verification, implementation, competent baselines, independent evaluation, and claim-specific transitions before support use. No local implementation, reproduction, performance, safety, deployment, support-state, or ASI result is established by this reconciliation row. |
87.15 Summary
The research agenda is the book’s backlog discipline. It keeps source ingestion, bibliography, proof targets, experiments, missing evidence, and chapter insertion rules visible so the living book can grow without losing evidence control.
After the exact core-claim mappings, the next bottleneck is no longer knowing which sources belong to which claims. It is doing the slower work: passage review, external citation normalization, artifact import or reproduction, and source-specific chapter revision without collapsing design rationale into unsupported evidence.
This final chapter is the handoff back to future work. The book can keep changing, but every change should leave a trace: what source moved, what chapter boundary changed, what claim state stayed the same or moved, what test or proof was affected, what edition was regenerated, and what remains blocked.
87.16 Evidence reconciliation (2026-07-16)
The invariant protocol, field meanings, and inference limits are stated once in Living Book Methodology. This packet contains only the chapter-specific projection; its authoritative per-atom rows are the open-research-agenda-and-bibliography-plan slice of experiments/claim_family_terminal_coverage/results/result.json.
The core remains blocked after full attempt at argument support. The strongest family attempt was Integrated governed lifecycle slices. Its exact boundary is: Bounded local replay only; no deployment, whole-book proof, external effect authority, transfer, publication, or release claim. Across 73 atoms, the terminal ledger records 73 blocked_after_full_attempt.
| Chapter-specific field | Value |
|---|---|
| Family / atom denominator | CF-08 / 73 atoms |
| Terminal dispositions | 73 blocked_after_full_attempt |
| Core | open-research-agenda-and-bibliography-plan.core: blocked_after_full_attempt at argument |
| Core attempted / missing lanes | causal, empirical, executable, formal, source-synthesis / normative, transfer |
| Attempted local lanes | causal, empirical, executable, formal, source-synthesis |
| Missing or unproved lanes | normative, transfer |
| Strongest family bundle | Integrated governed lifecycle slices (end_to_end): Three versioned integrated slices, 12 cases, all ten lifecycle states, six observed effects, rollback/residual/quarantine outcomes, and an eleven-surface sealed epoch. |
| Negative controls | 20 named boundary injections; three exact rollbacks; partial-effect residual and quarantine; 11 rejecting mutations. |
| Accepted transitions | v1_0_pilot.research_agenda.no_change |
| Maximum inference | Bounded local replay only; no deployment, whole-book proof, external effect authority, transfer, publication, or release claim. |
| Reproduction / next burden | Replay scripts/validate_p3_integrated_slices.py and scripts/validate_claim_family_terminal_program.py; fill the named atom-specific lanes under a new prospective protocol. |
87.17 Handoff
The book closes by returning its reader to the living workflow rather than to a sealed conclusion. The stack is now a governed research program: sources enter through inventory and notes, claims move only through evidence transitions, proofs stay narrow, tests and benchmarks require artifacts, reader editions derive from validated source, and every future change should leave a public-safe trace of what improved, what stayed speculative, and what remains blocked.