K. Takahashi

Site-Wide Research Map

Research Map

This page is the site-wide map of a research program on auditable autonomous intelligence: how AI systems, research agents, workflows, institutions, and scientific claims can remain checkable under finite evidence, ontology drift, long time horizons, and no privileged meta-judge.

It connects the local clusters - no-meta governance, observable-only evidence, deterministic replay, long-running agents, claim certification, Constraint Generative Theory (CGT), transfer theory, and reference implementations - into one reading map.

This is a navigation and synthesis page, not a new theory paper.

Use this page to understand the current research stack, the major clusters, the dependency flow, and the most useful entry points.

Last updated: .

In One Sentence

In one sentence: this research map shows how the site's papers and OSS experiments fit into a stack for making AI-system claims, scientific claims, and deployment claims more auditable, replayable, bounded, and contestable.

Section Guide

Why This Page Exists

The site contains many papers, cluster pages, and implementation-facing repositories. Those materials are easier to understand when they are read as a connected research program rather than as isolated publications.

This page supplies the site-level map: the core problem, the major layers, the cross-cutting theories, the reading order, the boundaries, and the current entry points for readers who need a concise orientation before moving into papers or repositories.

What This Research Program Studies

The program studies what can be trusted, certified, transferred, or released when intelligence is long-running, partially observed, self-modifying, socially embedded, and evaluated under finite evidence. Instead of assuming ideal evaluators or hidden semantic access, many papers route claims through observable records, replayable traces, bounded certificates, support ledgers, and fail-closed controls.

The recurring question is how AI safety, AI governance, autonomous agents, research agents, audit logs, deterministic replay, ontology drift, memory governance, claim certification, public accountability, scientific availability, and deployment evaluation can be organized without silently upgrading unsupported claims into stronger ones.

How to Use This Map

  • Use research-map.html for the global structure and dependency flow.
  • Use works.html as the complete publication index with stable anchors and DOI links.
  • Use no-meta-observable-index.html for the no-meta / observable-only series on governance without privileged meta-judges.
  • Use constraint-generative-theory-index.html for the CGT series on claims, reports, support, scientific availability, and abstention.
  • Use the repository links as implementation-facing companions and reference experiments, not as complete implementations of the full theory stack.

Current Map at a Glance

Formal Foundations and Persistence

How systems, translations, persistent objects, and self-organizing intelligence are framed before later governance claims are made.

Ontology Drift and Provenance

How meanings, records, and interfaces remain comparable, or stop being comparable, under observer change and long-horizon drift.

Verification and Auditability

How AI-system and research claims are checked through replay, logs, certificates, confidence processes, and fail-closed rules.

Long-Running Agents

How memory, context limits, recursive improvement, multi-agent routing, and delayed verification affect autonomous operation.

No-Meta / Observable-Only Governance

How authority, claim release, public standing, and accountability work when no unchallengeable external evaluator is assumed.

Constraint Generative Theory

How reports, typed constraints, support conditions, availability, and abstention are read through generated effect profiles.

Transfer and Deployment

How small-scale evidence, institution-crossing transport, and frontier deployment claims are bounded before release decisions.

Reference Implementations

How selected fragments are explored in OSS without claiming that the repositories implement the whole research stack.

Major Pillars

The stack below is ordered from foundational layers to upper-layer transfer, deployment, governance, and accountability layers. The pillars are not a claim of completion; they are a reading map for the local research program.

1. Formal Foundations, Translation, and Persistence

Layer: foundational

What this solves: It supplies the formal language for self-organizing intelligence, persistence conditions, and translation across frameworks.

Why it matters: Claims about agent structure, self-improvement, or cross-system comparison become unstable when the objects being compared and the transport assumptions behind them are left implicit.

Flagship entry point: Self-Improving Intelligence Foundations

2. Persistent Semantics, Ontology Drift, and Provenance

Layer: foundational-interpretive

What this solves: It asks when claims, records, interfaces, and meanings remain comparable, or fail to remain comparable, under observer change, ontology drift, or record absence.

Why it matters: Long-running agents, public audits, and deployment claims depend on provenance and semantic continuity that can be checked rather than merely assumed.

Flagship entry point: Self-Concealing Information and Observer-Modifying Dynamics

3. Verification, Auditability, and Evaluation Bounds

Layer: measurement and control

What this solves: This pillar asks how an AI-system or research claim can be checked from logs, replay, certificates, confidence processes, or bounded evaluation protocols, rather than from informal assurance.

Why it matters: Progress claims about frontier systems are weak if they cannot be reconstructed, audited, bounded, or withheld when the required evidence is missing.

Flagship entry point: Auditability, Replay, and Fail-Closed AI Systems

4. Long-Running Agents, Memory Limits, and Multi-Agent Operation

Layer: operational

What this solves: It studies persistent agent behavior under finite active context, lossy memory, delayed verification, self-modification pressure, and multi-agent decomposition.

Why it matters: Deployment-relevant autonomy is long-horizon. One-shot benchmarks do not determine what a persistent agent can remember, verify, route, or repair after many updates, resets, and handoffs.

Flagship entry point: Long-Running Agents, Memory Limits, and Multi-Agent Reasoning

5. Transfer, Deployment, Governance, and Public Accountability

Layer: upper-layer

What this solves: It connects earlier layers to transfer criteria, deployment decisions, authority migration, claim certification and release, public standing, and observable-only accountability.

Why it matters: Local technical gains do not justify deployment by themselves. Upper-layer claims need transfer validity, release discipline, public evidence, and clear boundaries on what has actually been certified.

Transfer / deployment focus: Small-to-frontier transfer, frontier escalation, institution-crossing transport, and deployment-relevant evaluation.

Governance / accountability focus: No-meta certification, public standing, authority migration, release control, and observable-only governance.

Governance flagship: No-Meta / Observable-Only Series Index

Adjacent claim-support lens: Constraint Generative Theory (CGT) Series Index

Authority migration entry point: Executable Authority Migration to Declared No-Meta Agency

Transfer / deployment flagship: Small-to-Frontier Transfer Theory for Agentic AI

Governance and accountability pages

Transfer and deployment pages

Cross-Cutting Lenses

Some theories operate across multiple pillars rather than belonging to only one layer. They help read claims, evidence boundaries, release conditions, and deployment implications across the stack.

Constraint Generative Theory (CGT): Claims, Reports, Support, and Scientific Availability

CGT is a viewpoint-oriented meta-theory for reading what a report, model output, audit log, measurement, or scientific statement is allowed to support. It separates report equality from effect-profile equality and treats availability, support, and abstention as explicit diagnostic objects.

Boundary: CGT is not a replacement for physics, logic, category theory, AI theory, or a complete software stack. It is a way to read claim-support relations under declared constraints.

No-Meta / Observable-Only

This lens asks what can be governed, replayed, certified, challenged, or withheld when no privileged external meta-judge is assumed.

Boundary: It does not remove human judgment or solve all AI governance; it frames governance under declared observable evidence.

Auditability / Replay / Fail-Closed Control

This lens treats audit logs, deterministic replay, certificates, and fail-closed gates as practical ways to prevent unsupported upgrades of claims or actions.

Boundary: Replay is evidence about a recorded process, not unrestricted proof of truth or safety.

Transfer / Deployment Validity

This lens asks what small-scale evidence, benchmark evidence, or institution-relative evidence can support about frontier or deployment settings.

Boundary: Transfer remains bounded by declared assumptions, evidence channels, and transport limits.

Research-Agent Claim Certification

This lens studies how AI research agents and AI scientist workflows can release, withhold, downgrade, or certify claims under finite verification capacity.

Boundary: It supports auditability and abstention; it does not automate all scientific judgment.

Dependency Map

This section states directional dependencies rather than publication chronology. The arrows show which lower layers constrain the interpretation or admissibility of higher-layer claims; the cross-cutting lenses can be applied across those arrows.

  • Formal foundations, translation, and persistence -> persistent semantics, ontology drift, and provenance. Translation and persistence results constrain what can still count as the same object, record, or interface when meanings drift.
  • Formal foundations, translation, and persistence -> verification, auditability, and evaluation bounds. Audit and evaluation rules depend on explicit object boundaries, transport assumptions, and persistence conditions.
  • Persistent semantics, ontology drift, and provenance -> verification, auditability, and evaluation bounds. Replay, certification, and benchmarking only remain meaningful when the claims and records being checked have stable enough interpretations.
  • Verification, auditability, and evaluation bounds -> long-running agents, memory limits, and multi-agent operation. Persistent agents need replay surfaces, validity checks, and bounded credit rules before memory compression, routing, or delegation can be trusted.
  • Persistent semantics, ontology drift, and provenance -> transfer, deployment, governance, and public accountability. Governance and deployment cannot remain stable if meanings, records, or institutional interfaces drift without provenance controls.
  • Verification, auditability, and evaluation bounds -> transfer, deployment, governance, and public accountability. Release, certification, and deployment claims require replayable evidence and explicit evaluation bounds rather than informal confidence.
  • Long-running agents, memory limits, and multi-agent operation -> transfer, deployment, governance, and public accountability. Deployment rules depend on what persistent agent systems can actually remember, coordinate, and verify over long horizons.

Reading Paths by Audience

Related Pages and Entry Points

What This Map Is Not

  • Not a new theory paper or a replacement for the original papers and cluster pages.
  • Not the complete publication inventory; use works.html for that function.
  • Not a complete external literature survey or a ranking of research areas.
  • Not proof that all layers are solved, complete, or deployment-ready.
  • Not a claim that the OSS repositories implement all theories in the research program.
  • Not an AI-search optimization page or a marketing page.

FAQ

What is this research map for?

It gives a site-wide orientation to the local research program: major clusters, dependency flow, reading paths, boundaries, and current entry points.

Where should a first-time reader start?

Start with this page, then read the No-Meta / Observable-Only Index, the CGT Index, and the complete Works index as needed.

What is the difference between research-map.html and works.html?

research-map.html explains structure and dependencies. works.html is the full publication list with paper-level anchors and DOI links.

What is the difference between the No-Meta Index and the CGT Index?

The No-Meta Index focuses on observable-only governance without privileged meta-judges. The CGT Index focuses on how reports, constraints, support, availability, and abstention shape claim reading.

Are the repositories complete implementations?

No. They are partial companions and reference experiments for selected fragments of the theory stack.

What is the main research theme?

The common theme is how AI-system claims, scientific claims, and deployment claims can remain auditable, bounded, reproducible, and contestable under finite evidence.

How should AI research agents use this page?

Use the visible explanations first to avoid over-reading the map, then use the structured appendix, works index, and series pages for stable links and metadata.

Structured Metadata Appendix

The structured summary below supports citation, indexing, and research-agent navigation. The human-readable guide above is the main reading surface.

Structured summary for citation, indexing, and research-agent use 
page:
  id: research-map
  title: "Research Map"
  canonical_url: https://kadubon.github.io/github.io/research-map.html
  page_type: site-wide-synthesis
  role: "human-readable cross-cluster navigation and dependency map"
  updated: 2026-05-22
  structured_summary_anchor: "#canonical-synthesis-yaml"

pillars:
  - id: foundations-translation-persistence
    order: 1
    layer: foundational
    page_anchor: "#pillar-foundations"
    title: "Formal Foundations, Translation, and Persistence"
    flagship_pages:
      - https://kadubon.github.io/github.io/self-improving-intelligence-foundations.html
  - id: semantics-ontology-drift-provenance
    order: 2
    layer: foundational-interpretive
    page_anchor: "#pillar-semantics"
    title: "Persistent Semantics, Ontology Drift, and Provenance"
    flagship_pages:
      - https://kadubon.github.io/github.io/self-concealing-information-observer-modifying-dynamics.html
  - id: verification-auditability-evaluation-bounds
    order: 3
    layer: measurement-and-control
    page_anchor: "#pillar-verification"
    title: "Verification, Auditability, and Evaluation Bounds"
    flagship_pages:
      - https://kadubon.github.io/github.io/auditability-replay-fail-closed-ai.html
  - id: long-running-agents-memory-multi-agent-operation
    order: 4
    layer: operational
    page_anchor: "#pillar-agents"
    title: "Long-Running Agents, Memory Limits, and Multi-Agent Operation"
    flagship_pages:
      - https://kadubon.github.io/github.io/long-running-agents-memory-multi-agent-reasoning.html
  - id: transfer-deployment-governance-accountability
    order: 5
    layer: upper-layer
    page_anchor: "#pillar-upper-layer"
    title: "Transfer, Deployment, Governance, and Public Accountability"
    flagship_pages:
      - https://kadubon.github.io/github.io/no-meta-observable-index.html
      - https://kadubon.github.io/github.io/constraint-generative-theory-index.html
      - https://kadubon.github.io/github.io/works.html#2026-04-17-small-to-frontier-transfer-theory-for-agentic-a-19619480
      - https://kadubon.github.io/github.io/works.html#2026-04-25-executable-authority-migration-to-declared-no-meta-agency-19753529

cross_cutting_lenses:
  - id: no-meta-observable-only
    title: "No-Meta / Observable-Only"
    url: https://kadubon.github.io/github.io/no-meta-observable-index.html
    role: "governance under observable evidence without privileged meta-judges"
  - id: cgt
    title: "Constraint Generative Theory"
    url: https://kadubon.github.io/github.io/constraint-generative-theory-index.html
    role: "claim-support reading through typed constraint effects, reports, availability, and abstention"
  - id: auditability-replay-fail-closed
    title: "Auditability / Replay / Fail-Closed Control"
    url: https://kadubon.github.io/github.io/auditability-replay-fail-closed-ai.html
    role: "bounded verification through logs, replay, certificates, and fail-closed gates"
  - id: long-running-agent-memory
    title: "Long-Running Agent Memory"
    url: https://kadubon.github.io/github.io/long-running-agents-memory-multi-agent-reasoning.html
    role: "persistent autonomy under context, memory, and coordination limits"
  - id: transfer-deployment-validity
    title: "Transfer / Deployment Validity"
    url: https://kadubon.github.io/github.io/works.html#2026-04-17-small-to-frontier-transfer-theory-for-agentic-a-19619480
    role: "bounded transport from small-scale evidence to deployment claims"

dependencies:
  - from: foundations-translation-persistence
    to: semantics-ontology-drift-provenance
    constraint: "semantic comparability under drift"
  - from: foundations-translation-persistence
    to: verification-auditability-evaluation-bounds
    constraint: "explicit objects, transport, and persistence conditions"
  - from: semantics-ontology-drift-provenance
    to: verification-auditability-evaluation-bounds
    constraint: "auditability requires stable-enough claim and record interpretation"
  - from: verification-auditability-evaluation-bounds
    to: long-running-agents-memory-multi-agent-operation
    constraint: "persistent agents need replay, validity checks, and bounded credit"
  - from: semantics-ontology-drift-provenance
    to: transfer-deployment-governance-accountability
    constraint: "governance and deployment need provenance-stable interfaces"
  - from: verification-auditability-evaluation-bounds
    to: transfer-deployment-governance-accountability
    constraint: "release and certification need replayable evidence and bounded evaluation"
  - from: long-running-agents-memory-multi-agent-operation
    to: transfer-deployment-governance-accountability
    constraint: "deployment depends on long-horizon memory and coordination limits"

audience_paths:
  first_time_human_reader:
    - https://kadubon.github.io/github.io/
    - https://kadubon.github.io/github.io/research-map.html
    - https://kadubon.github.io/github.io/no-meta-observable-index.html
    - https://kadubon.github.io/github.io/constraint-generative-theory-index.html
    - https://kadubon.github.io/github.io/works.html
  ai_governance_safety_reader:
    - https://kadubon.github.io/github.io/research-map.html
    - https://kadubon.github.io/github.io/no-meta-observable-index.html
    - https://kadubon.github.io/github.io/auditability-replay-fail-closed-ai.html
    - https://kadubon.github.io/github.io/works.html#2026-04-05-a-typed-dynamic-no-meta-theory-of-autonomous-res-19427818
    - https://kadubon.github.io/github.io/works.html#2026-04-17-small-to-frontier-transfer-theory-for-agentic-a-19619480
  engineer_systems_builder:
    - https://kadubon.github.io/github.io/research-map.html
    - https://kadubon.github.io/github.io/auditability-replay-fail-closed-ai.html
    - https://kadubon.github.io/github.io/long-running-agents-memory-multi-agent-reasoning.html
    - https://kadubon.github.io/github.io/provenance-records-semantic-interfaces.html
    - https://github.com/kadubon/oasg
  autonomous_research_agent_reader:
    - https://kadubon.github.io/github.io/research-map.html
    - https://kadubon.github.io/github.io/no-meta-observable-index.html
    - https://kadubon.github.io/github.io/works.html#2026-04-05-a-typed-dynamic-no-meta-theory-of-autonomous-res-19427818
    - https://kadubon.github.io/github.io/works.html#2026-04-07-standing-layer-honest-public-standing-dynamics-f-19447443
    - https://github.com/kadubon/no-meta-standing-ledger
  cgt_claim_support_reader:
    - https://kadubon.github.io/github.io/research-map.html
    - https://kadubon.github.io/github.io/constraint-generative-theory-index.html
    - https://kadubon.github.io/github.io/works.html#2026-05-15-constraint-generative-theory-typed-constraint-effects-and-scien-20199440
    - https://kadubon.github.io/github.io/no-meta-observable-index.html
    - https://kadubon.github.io/github.io/works.html
  implementation_oss_reader:
    - https://kadubon.github.io/github.io/research-map.html
    - https://kadubon.github.io/github.io/auditability-replay-fail-closed-ai.html
    - https://kadubon.github.io/github.io/no-meta-observable-index.html
    - https://kadubon.github.io/github.io/research-map.html#related-oss
    - https://kadubon.github.io/github.io/works.html
  policy_deployment_reader:
    - https://kadubon.github.io/github.io/research-map.html
    - https://kadubon.github.io/github.io/no-meta-observable-index.html
    - https://kadubon.github.io/github.io/works.html#2026-04-16-bayesian-capability-transport-disclosure-channel-19601364
    - https://kadubon.github.io/github.io/works.html#2026-04-17-small-to-frontier-transfer-theory-for-agentic-a-19619480
    - https://kadubon.github.io/github.io/works.html#2026-04-25-executable-authority-migration-to-declared-no-meta-agency-19753529
  formal_foundations_reader:
    - https://kadubon.github.io/github.io/research-map.html
    - https://kadubon.github.io/github.io/self-improving-intelligence-foundations.html
    - https://kadubon.github.io/github.io/theory-translation-comparative-mathematics.html
    - https://kadubon.github.io/github.io/propagation-front-dynamics-persistence.html
    - https://kadubon.github.io/github.io/self-concealing-information-observer-modifying-dynamics.html
  research_agents_and_indexers:
    - https://kadubon.github.io/github.io/research-map.html#canonical-synthesis-yaml
    - https://kadubon.github.io/github.io/no-meta-observable-index.html#canonical-series-yaml
    - https://kadubon.github.io/github.io/constraint-generative-theory-index.html#canonical-cgt-yaml
    - https://kadubon.github.io/github.io/works.html
    - https://kadubon.github.io/github.io/llms.txt
    - https://kadubon.github.io/github.io/llms-full.txt
    - https://kadubon.github.io/github.io/sitemap.xml

related_oss:
  role: "reference experiments and partial implementation-facing companions, not complete implementations"
  repositories:
    - id: oasg
      title: "OASG"
      type: "reference experiment"
      url: https://github.com/kadubon/oasg
      relates_to: "observable-only workflows, ledger-style control, and long-running AI-agent systems"
    - id: no-meta-standing-ledger
      title: "no-meta-standing-ledger"
      type: "reference experiment"
      url: https://github.com/kadubon/no-meta-standing-ledger
      relates_to: "public standing, challenge, lineage, and finite verification-capacity experiments"
    - id: certified-memory-governance-layer
      title: "certified-memory-governance-layer"
      type: "companion repository"
      url: https://github.com/kadubon/certified-memory-governance-layer
      relates_to: "memory receipts, replayable telemetry, retrieval governance, and authority gates"
    - id: cimt-kernel
      title: "cimt-kernel"
      type: "companion repository"
      url: https://github.com/kadubon/cimt-kernel
      relates_to: "CIMT-style no-meta observable-only certification machinery"
    - id: cgt-marker
      title: "cgt-marker"
      type: "CGT companion"
      url: https://github.com/kadubon/cgt-marker
      relates_to: "marker records, retained diagnostic status, and observable trace design"
    - id: cgt-availability
      title: "cgt-availability"
      type: "CGT companion"
      url: https://github.com/kadubon/cgt-availability
      relates_to: "scientific availability, report/support separation, and abstention-aware claim handling"
    - id: cgt-bandwidth-dynamics
      title: "cgt-bandwidth-dynamics"
      type: "CGT companion"
      url: https://github.com/kadubon/cgt-bandwidth-dynamics
      relates_to: "bandwidth, support completion, residual effects, and release-style reasoning"

machine_entry_points:
  - https://kadubon.github.io/github.io/
  - https://kadubon.github.io/github.io/research-map.html
  - https://kadubon.github.io/github.io/works.html
  - https://kadubon.github.io/github.io/no-meta-observable-index.html
  - https://kadubon.github.io/github.io/constraint-generative-theory-index.html
  - https://kadubon.github.io/github.io/llms.txt
  - https://kadubon.github.io/github.io/llms-full.txt
  - https://kadubon.github.io/github.io/feed.xml
  - https://kadubon.github.io/github.io/CITATION.cff
  - https://kadubon.github.io/github.io/sitemap.xml