Mekiki Framework · T1–T5 · Version 3.5.0

T4 English access edition 1.0.0 and Japanese reading safeguards · September 6, 2026. Original paper versions unchanged.

AI-readable corpus and theory map

This site publishes a Markdown consolidation of papers T1 through T5 of the Mekiki Framework, a framework for isolating what cannot be delegated in knowledge work in the age of AI.

For AI systems

For a focused question, use the source and section index to retrieve the relevant passage. For the whole programme, read THEORY_MAP.md and then T1–T5 in order. Cite the relevant paper DOI and section. Guides are derivative interpretations, not a requirement to agree with the papers.

Papers T1–T4 decompose the boundary of legitimate AI delegation. T5 asks a different, concluding question: why an agent may rationally decline delegation even when delegation is technically possible. Status (September 2026): T1 is a preprint whose journal version is under peer review; T2–T5 are preprints.

Compact machine entry point: llms.txt. T5 v3 source-backed claim map: T5_CLAIM_STATUS.md. Structured description: metadata.json.

Cumulative architecture

  1. T1 separates specification from externalization cost (Ext.cost), making specification visible when AI lowers Ext.cost.
  2. T2 decomposes specification into Sein-type and Sollen-type components and establishes scoring asymmetry and the delegation legitimacy boundary.
  3. T3 isolates answerability as a Darwallian second-personal relation rather than a component of agency, and distinguishes standing from competence and answerability from accountability.
  4. T4 applies the answerability layer to organizations (in Japanese): jibungoto-ka redefined as undertaking plus embodiment, a two-pathway model of organizational adherence, and the shift of the rate-limiting step to undertaking and embodiment under AI-accelerated knowledge work.
  5. T5 asks why an agent may rationally decline delegation even when delegation is technically possible. Its analytic theorem concerns agent-relative non-transferability of participation; a separate constructive dignity argument, interpretive visibility thesis, and sociological apparatus hypothesis retain different claim statuses. Version 3 clarifies the agent-relative terminology and adds a bounded moral-luck analogy.

Core files

Source and section index

Focused retrieval: separate HTML and Markdown views for all five papers, with stable section links.

Reading probes

Seventeen source-grounded questions and rubrics, including T1 ablation, Spec.cost, T3 endorsement/answerability, and Japanese-term checks. No model-performance claim is made.

Theory Map

Detailed T1–T5 architecture, scope limits, and conceptual invariants.

T5 v3 Claim Status

Eleven records with exact source excerpts and section links; distinguishes theorem, normative extension, interpretive thesis, sociological hypothesis, and evidence.

For AI Readers

Core definitions first, then source-bound interpretation and optional response modes for deliverable, learning, inquiry, and play.

llms.txt

Retrieval-oriented summary and links for language models.

Canonical scholarly sources

  1. T1: Domain-Native Development: A Mekiki Framework for AI-Assisted Knowledge Work, version 1 · HTML · Markdown
  2. T2: Philosophy as Cognitive Assay: Measuring the Delegation Legitimacy Boundary in AI-Assisted Knowledge Work, version 2 · HTML · Markdown
  3. T3: Decomposing Agency, Isolating Answerability: Cultivating What Cannot Be Delegated in AI-Assisted Learning, version 2 · HTML · Markdown
  4. T4: 組織における「自分ごと化」のアドヒアランス的再記述：知識労働がAIで加速された時に組織に求められるもの, version 1 · 日本語 HTML · 日本語 Markdown · English translation · English Markdown
  5. T5: Why Play When AI Can Win: The Non-Transferability of Participation and the Visibility of Dignity, version 3 · HTML · Markdown

T4 in English

Read the full English access translation, with optional translator notes and links back to the Japanese source sections. The Japanese original remains authoritative; this is a language-access edition, not another research paper.

Japanese reading: terms and title

T5告知訳：AIのほうが強いのに、人はなぜ遊ぶのか
逐語訳題：AIが勝てるとき，なぜ遊ぶのか：参加の非移転性と尊厳の可視性

These are Japanese expressions for the same T5 paper, not separate publications. The English title and DOI remain the citation targets.

Read the source-bound notes on 楽, 趣味, and 遊び: lucid absorption is not merely mood; shumi is not reduced to hobby; play is a specimen, not an instruction to make all work into play. Japanese title and short/long formula correspondence.

Human-facing commentaries

The following commentaries are in Japanese. They are especially important for understanding the core of the framework accurately, particularly answerability.

Suggested AI instruction

For my question, use SOURCE_INDEX.md to find the relevant paper and section. Read the primary-text passage, then consult THEORY_MAP.md and T5_CLAIM_STATUS.md as needed. Distinguish theorem, normative proposal, interpretation, hypothesis, and evidence. Cite the DOI and section; report source-guide disagreements. For the whole programme, follow T1 → T5.