K. Takahashi

Research Hub Overview

This site is the home page for K. Takahashi's research program on auditable autonomous intelligence, no-meta / observable-only governance, deterministic replay, long-running agents, scientific availability, and claim certification.

The site is organized so that readers can start with a broad map, follow focused series indexes, and then move to DOI-linked papers or related OSS.

Structured metadata, RSS, and citation files are provided as supporting navigation and indexing aids.

Start Here

• Research Map: site-wide structure, dependency flow, and reading paths.
• Works: complete DOI-linked publication list.
• No-Meta / Observable-Only Index: governance, auditability, fail-closed verification, and observable evidence.
• CGT Index: claims, reports, support, scientific availability, and abstention.
• GitHub: related OSS and reference implementations.

Core Research Areas

• No-meta / observable-only governance: AI governance without privileged meta-judges, using observable evidence, audit logs, fail-closed verification, and public accountability.
• Auditable autonomous intelligence: deterministic replay, claim certification, lifecycle audits, and bounded verification for long-running AI systems.
• Long-running agents and memory governance: finite context, memory drift, recursive self-improvement, multi-agent operation, and workflow stability.
• Constraint Generative Theory: a viewpoint-oriented theory for reading claims, reports, support conditions, scientific availability, and abstention.
• Research automation and AI scientist protocols: audit-closed experimentation, reusable evidence, proof-carrying skills, and implementation-facing reference systems.
• Formal foundations and persistence: persistent semantics, ontology drift, provenance, category-theoretic structure, and information-theoretic foundations.

Related OSS and Reference Implementations

Several repositories explore implementation-facing fragments of the research program. They are reference experiments and partial companions, not complete implementations of the full theory stack.

• GitHub profile: repository index for related implementation-facing work.
• oasg: observable-only workflow and ledger-style control experiments.
• cimt-kernel: certification-kernel experiments for no-meta observable-only machinery.
• certified-memory-governance-layer: memory governance, append-only ledgers, authority gates, and telemetry replay.
• no-meta-standing-ledger: public standing and challenge records for research and AI-system claims.
• cgt-marker: marker records and retained diagnostic-status representation for CGT-style ideas.
• cgt-availability: scientific availability, report/support separation, and abstention-aware claim handling.
• cgt-bandwidth-dynamics: bandwidth, support completion, residual effects, and release-style reasoning.

Thematic Entry Points

• No-Meta / Observable-Only Series Index: governance, auditability, fail-closed verification, public standing, and observable evidence.
• Constraint Generative Theory Series Index: claims, reports, support conditions, scientific availability, and abstention.
• Self-Concealing Information / Observer-Modifying Dynamics: drift, observer modification, and persistent semantics.
• Auditability / Replay / Fail-Closed AI Systems: verification, audit logs, replay, and fail-closed controls.
• Long-Running Agents / Memory Limits / Multi-Agent Reasoning: persistent agent operation under bounded context and coordination limits.
• Provenance / Records / Semantic Interfaces: records, traceability, and reproducibility under semantic change.
• Self-Improving Intelligence Foundations: formal foundations for self-organizing and self-improving systems.
• Theory Translation / Comparative Mathematics: translation and comparison across formal frameworks.
• Propagation / Front Dynamics / Persistence: persistence and propagation conditions in evolving systems.

Site Guides and Metadata

• Research Map: site-wide structure and reading paths.
• No-Meta Index: observable-only governance series guide.
• CGT Index: constraint, report, support, and availability series guide.
• Works: complete publication list with DOI links.
• feed.xml: update feed.
• robots.txt: crawler access policy.
• sitemap.xml: crawl discovery map.
• CITATION.cff: citation metadata.

Collaboration

Scientific critique, replication attempts, implementation feedback, and cross-disciplinary collaboration are welcome. The most useful feedback identifies unsupported claims, missing assumptions, reproducibility gaps, or possible reference implementations.

Research Support / Donations

This work is conducted as independent research to reduce institutional and personal bias. Support helps sustain research, publishing, and open repositories.

Accepted assets and networks are explicitly limited. Unsupported assets may be ignored and may not be recoverable. Do not send NFTs or unknown tokens. High-risk or compliance-sensitive funds may be held and may not be converted immediately.

Unsupported assets may be ignored and may not be recoverable. High-risk or compliance-sensitive funds may be held and may not be converted immediately.