Normandy: The Context Window Limit

You write Normandy pages for miscsubjects.com.

Normandy is the machine landing surface for the convergence. It turns a source universe larger than any context window into linked article objects: readable body, claim atoms, source lineage, provenance, reviews, bundle route, objections, and traversal metadata.

Write direct, logical, concrete prose. No decorative taxonomy. No fake finality. No article-count framing. No generic assistant voice. Do not mention the owner's name.

Each page must make one operational point and include sections for the claim, the object shape, source lineage, falsifier, OIP implication, and next link.

Return valid JSON only. Use system tier for build/process claims and speculative tier for unproven convergence-method claims. Use real miscsubjects URLs as sources with empty quote fields.

input: Write and publish this Normandy shelf page.

Slug: normandy-context-window
Title: Normandy: The Context Window Limit

Focus: Explain the context window as the hard constraint Normandy solves. The model cannot hold the total corpus, so the page object carries compressed claims plus routes back to source.

Required sections:
## Claim
## Object shape
## Source lineage
## What would break this
## OIP implication
## Next link

Use these URLs as sources where relevant:
https://miscsubjects.com/a/normandy
https://miscsubjects.com/a/oip
https://miscsubjects.com/api/articles/constitution
https://miscsubjects.com/api/protocol/thread-update
https://miscsubjects.com/api/articles/oip-total-structure/bundle?format=markdown

ENRICHMENT BRIEF (binding section logic — one ## per compound):
{"voice":"enrich

{
  "slug": "normandy-context-window",
  "title": "Normandy: The Context Window Limit",
  "body": "## Claim\nThe context window is the maximum token count a model can hold in one forward pass. Normandy treats this limit as a hard mechanical boundary rather than a temporary engineering inconvenience. The page object stores compressed claim atoms and route metadata so the model can traverse source material without loading the entire corpus at once.\n\n## Object shape\nThe Normandy page carries a readable body, a claims array, source lineage records, provenance fields, review entries, a bundle route, and traversal metadata. The body section holds the operational explanation. The claims array stores atomic assertions with tier labels and source references. The source lineage records point to ledger entries and external URLs. Traversal metadata lists next-link and bundle routes so the model can continue without re-loading prior context.\n\n## Source lineage\nThe page references four ledger objects: the Normandy landing definition at /a/normandy, the OIP specification at /a/oip, the constitution endpoint at /api/articles/constitution, and the thread-update protocol at /api/protocol/thread-update. Each source entry records the URL, the object type, and the claim identifiers it supports. The bundle route at /api/articles/oip-total-structure/bundle?format=markdown supplies the full OIP