chronicle · method note

HOW THIS CHRONICLE VERIFIES

the pipeline behind every cited line
● method page — 2026-09-04

the diagram

Every claim in the Manipur Chronicle passes through the same pipeline — fetched, dated, quoted, cross-checked — or it is labeled. This page is the pipeline, drawn.

Flow diagram: how the Manipur Chronicle verifies a claim — nine steps from story on the field to publish, with a two-sources decision point, a single-source caution box, and notes on numbers conflict, the citation ledger, and the Wikipedia second eye.
the verification pipeline. editable source: method-diagram.excalidraw (drag it into excalidraw.com) · static copy: method-diagram.svg

the rules, in order

  1. Story on the field. A claim is collected from news pages, Wikipedia (including talk pages), or social channels.
  2. Fetch-once guard. Each page is fetched once and cached in a manifest; it is never re-fetched without a logged reason.
  3. Dated, not just live. Archive capture-date (CDX) windows confirm a page said X at the time of the event — not only today.
  4. Ledger + verbatim. Every source becomes a numbered citation carrying a verbatim quote that must literally appear in the fetched text. A paraphrase cannot pass the check.
  5. Two or more sources? If a story has 2+ sources, the cluster is tone-scanned — framing terms per outlet — and recorded in a story-cluster file. A single source is quoted and explicitly labeled a lone-source claim.
  6. Tone-scan the cluster. Headline verbs, labels, whose quotes carried, what context was added or omitted — compared outlet by outlet.
  7. Multi-angle check. Wikipedia talk pages are read for edit wars, reverts, and neutrality disputes; outlet lean and retractions are noted.
  8. Perception layer. Social-media content is perception data, never factual evidence — and bot-like amplification, when detected, is itself a documented finding.
  9. Publish. Conflicting numbers appear side by side, never merged. Claims that cannot be sourced are flagged [unverified]. Every installment ends with a bias & provenance watch.

Stranger-safe throughout: no internal system detail, no victim names beyond published sources, no rumor amplification except as labeled examples of the disinformation layer.

limits, honestly

The method verifies what was published and when — it cannot verify what happened offline. Sources can also be wrong in the same direction, which is why conflicting readings are never merged and why the chronicle reports claims about events rather than certifying the events themselves.

Gaps are logged, not hidden: open questions appear in the installments as future work, and the monthly timeline runs backfill items when a fresh archive window surfaces them.