Skip to content

Summary

This bundle organises research on the political landscape of Japanese mass media, with a specific focus on how left-leaning voices are represented, and a forward-looking section on what it would take to launch a liberal-leaning media outlet in Japan. It exists so that a human and an LLM agent can extend the same body of work without re-deriving it, and so that every claim carries its provenance status.

It is a synthesis layer: concepts distil and cross-link, they do not restate each other. Evidence lives in references/evidence-ledger.md as numbered sources; concept bodies cite them by footnote.

Scope

In scope: the structure of the Japanese media system; the left–right positioning of outlets; the mechanisms by which left-leaning voices are marginalised or framed; the named actors and their documented overlaps; and the practical constraints on starting a new liberal-leaning outlet.

Out of scope: general Japanese politics except where it bears on media; US/European media comparisons beyond a single sentence of context; anything requiring a paid subscription that we have not read (state the gap instead).

The Claim-Status Policy (this bundle's core discipline)

This is a politically contested domain, and the failure mode is not missing facts — it is laundering inference into fact. Every substantive claim carries one of three labels, written inline in the sentence or the bullet:

  • documented — a named actor, a primary record, or an authoritative institutional source. Cite the source id.
  • inferred — consistent with the documented record but not proven as stated. Say what would confirm or refute it.
  • contested — asserted by an interested party (a partisan outlet, a state media organ, an advocacy group) and not independently established. Name the interest.

Rules:

  1. A claim that is inferred may never be cited downstream as documented.
  2. A contested claim may be reported (that X says Y) but never asserted (that Y is true).
  3. Absence of evidence is recorded as a coverage gap, not as evidence of absence.
  4. Aggregation is not corroboration: ten copies of one wire story are one source.

Authoring Conventions (OKF v0.2)

Frontmatter template for every concept — see templates/concept.md:

---
type: Concept        # Concept | Topic | Playbook | Case Study | Reference
title: <display name>
description: <one line>
tags: [..]
generated: { by: <agent>/<version>, at: <ISO-8601 UTC> }
status: draft        # draft | stable | deprecated
stale_after: <ISO-8601 UTC>   # required for 95-inbox items and dated playbooks
sources:
  - id: <short-id>
    resource: <bundle-relative path | URL>
    title: <label>
---
  • Internal links are bundle-rooted: [Kisha clubs](00-foundations/kisha-club-system.md).
  • Body shape: # Summary (2–4 sentences) → 2–5 content sections → # Links → # Footnotes.
  • Per-claim attribution uses markdown footnotes keyed to sources[].id.
  • Target 60–120 lines per concept. If a concept is outgrowing that, split it.

Status and Promotion

  • Every concept created by an agent is status: draft.
  • Promotion to stable is a human:yasu verified event — add verified: { by: human:yasu, at: <ISO-8601 UTC> }. Agents never self-promote.
  • Nothing in this bundle is stable yet.

News-Intake Policy

  • Current-events items enter via 95-inbox as append-only, dated concepts, filename YYYY-MM-DD-<slug>.md.
  • Required metadata: published, retrieved, as_of, canonical URL, and a snapshot note (quote or hash) so a revised page cannot masquerade as the version we reasoned from.
  • Every inbox concept carries stale_after. Expired items become status: deprecated, never deleted.
  • Intake convention: 95-inbox/source-intake-convention.md.

Expansion Workflow

To add a concept: python3 tools/new_concept.py <holder> <slug> --title "...", then fill it in and add a line to that holder's index.md. Run python3 tools/validate_bundle.py before committing. The full procedure is in 90-playbooks/bundle-expansion-workflow.md.

Source Notes


  1. The conventions follow the working OKF v0.2 bundle at ../options-system-wiki/, which is the local reference implementation. The standalone spec file was not present on this host at authoring time. ↩

  2. The ledger is a mechanical export of the retrieval-time citation ledger; ids are stable within this bundle. ↩