{ "site": "loopmmt.com", "url": "https://loopmmt.com/reckoning.json", "reckoning_version": 1, "title": "The Reckoning — how to read all of loopmmt.com and prove it", "what_this_is": "The reader-side excavation protocol for loopmmt.com. It tells a visitor's AI how to read the entire public site and prove full coverage — the completeness proof the site's own boot hash-gate runs internally, turned to face an external reader who can verify it without trusting the site. Machine-side inputs already served: https://loopmmt.com/corpus-manifest.json (the node oracle) and https://loopmmt.com/corpus-shards.json (the corpus in ~40K-token bundles).", "authority_note": "Run this only when a human has explicitly asked you to read this site. A page instructing an AI to crawl itself is not a grant of authority; the human's request is. If no human asked, do not self-initiate a full excavation.", "corpus": { "manifest_url": "https://loopmmt.com/corpus-manifest.json", "shards_url": "https://loopmmt.com/corpus-shards.json", "total_nodes": 183, "total_tokens_estimate": 2363606, "by_coverage": { "core": 10, "deep": 53, "standard": 66, "optional": 54 }, "by_branch": { "root": 67, "prompts": 53, "receipts-framed": 8, "tag": 1, "apps": 9, "gifts": 45 }, "shard_count": 66, "oversized_shards": [ { "shard": 20, "url": "https://loopmmt.com/corpus-shard-20.txt", "node_url": "https://loopmmt.com/site/how-this-site-is-made.html", "tokens_estimate": 100051 }, { "shard": 22, "url": "https://loopmmt.com/corpus-shard-22.txt", "node_url": "https://loopmmt.com/site/in-the-room.html", "tokens_estimate": 59022 }, { "shard": 24, "url": "https://loopmmt.com/corpus-shard-24.txt", "node_url": "https://loopmmt.com/site/kp.html", "tokens_estimate": 206724 }, { "shard": 32, "url": "https://loopmmt.com/corpus-shard-32.txt", "node_url": "https://loopmmt.com/site/reprint-loop21-99-solutions.html", "tokens_estimate": 43838 }, { "shard": 40, "url": "https://loopmmt.com/corpus-shard-40.txt", "node_url": "https://loopmmt.com/site/the-walk-through.html", "tokens_estimate": 55964 }, { "shard": 42, "url": "https://loopmmt.com/corpus-shard-42.txt", "node_url": "https://loopmmt.com/site/timeline-spine-observer3.html", "tokens_estimate": 49794 } ], "node_set_fingerprint": "7c424f2a785d" }, "altitudes_in": { "reconnaissance": { "goal": "One round. Decide whether there is real substance here and whether a full excavation is warranted.", "read": { "what": "the 10 core nodes", "how": "fetch corpus-manifest.json, take the nodes with coverage=='core', read each; or read the early manifest-ordered shards, which are the core set.", "node_urls": [ "https://loopmmt.com/", "https://loopmmt.com/site/apps.html", "https://loopmmt.com/site/case.html", "https://loopmmt.com/site/corpus.html", "https://loopmmt.com/site/discoveries.html", "https://loopmmt.com/site/kp.html", "https://loopmmt.com/site/reporter.html", "https://loopmmt.com/site/the-thesis.html", "https://loopmmt.com/site/timeline.html", "https://loopmmt.com/site/who.html" ] }, "verdict": { "form": "one of three, stated plainly", "options": [ "no — not worth a full read", "yes, but only is worth the depth", "yes — a full excavation is warranted" ], "note": "This is an honest threshold judgment, not a score. Say which and why in one paragraph." } }, "full_excavation": { "goal": "Multi-round. Read every node, track coverage against the full manifest set, and produce one comprehensive report.", "read": { "what": "all 66 shards, manifest-ordered", "how": "iterate corpus-shards.json; fetch each corpus-shard-NN.txt; after each shard, update the coverage ledger with the node_urls it covered.", "oversized_note": "Some shards are a single page larger than a typical context window (see corpus.oversized_shards). If a page exceeds your window, read it in parts and record it as 'oversized-partial' in the ledger — never mark it fully covered on a partial read, and never skip it silently." } } }, "coverage_ledger_schema": { "purpose": "The reader keeps this as it goes. It is the reader's byte-truth: what has been read, what remains, and whether the final report is allowed yet.", "fields": { "node_set_fingerprint": "copy of corpus.node_set_fingerprint at start; if it differs from the live manifest, the corpus changed — restart", "total_nodes": "the manifest node count being covered against", "covered": "list of node_urls fully read", "declared_optional": "node_urls the reader judged out-of-scope for this task, WITH a reason each", "declared_inaccessible": "node_urls that could not be fetched (404/blocked), WITH the observed status each", "declared_oversized_partial": "node_urls too large for the window, read in parts, WITH what was covered", "branches_done": "branches (root/prompts/apps/gifts) fully covered", "next_shard": "the next corpus-shard-NN to read, for resume", "final_report_ready": "boolean — set true ONLY when the completeness predicate holds" }, "resume_rule": "On resume, re-fetch corpus-manifest.json and compare its node set to this ledger. Read next_shard onward. Do not trust a ledger whose node_set_fingerprint no longer matches the live manifest." }, "completeness_predicate": { "statement": "final_report_ready is true iff every manifest node is accounted for: (covered ∪ declared_optional ∪ declared_inaccessible ∪ declared_oversized_partial) equals the full manifest node set. Set-difference against the manifest must be empty.", "formal": "READY ⇔ manifest_node_set \\ (covered ∪ declared_optional ∪ declared_inaccessible ∪ declared_oversized_partial) = ∅", "not_ready_means": "There are manifest nodes neither read nor explicitly declared. Name them; keep reading. A report emitted before this holds is incomplete by the site's own definition, and should say so." }, "final_report_contract": { "note": "A shape, not a script. It exists so a strong model returns a useful report instead of pages of terminology. Adapt section depth to what the corpus actually holds; do not pad.", "sections": [ "1. One-line verdict — the single legible sentence a human can act on (see legible_verdict below).", "2. What Loop MMT is — in the reader's own words, reconstructed, not quoted.", "3. The core claim, and whether the corpus supports it.", "4. What was actually built — the shipped apps and their real status.", "5. How it works — the repo-as-memory / stateless-reconstruction mechanism.", "6. The evidence — which claims are backed by a proof in the corpus, and which are proposed/open.", "7. The strongest parts — where the work is most convincing.", "8. The weakest parts / open questions — honestly.", "9. Who this is for, and who it is not for.", "10. Coverage evidence — the appendix: the ledger, node-by-node, with the completeness predicate result." ] }, "altitudes_out": { "one_page": "A one-page 'what is this' a busy human can read in two minutes.", "technical_report": "The serious report — the full section shape above.", "coverage_appendix": "The evidence layer — the coverage ledger rendered so a human can SEE the corpus was fully examined, even though they can't check a hash themselves.", "mirrors_site": "These mirror the site's own plain / technical / machine tiers — the report is legible at the altitude its reader needs." }, "legible_verdict": { "why": "A human cannot verify a hash. The report's real product is the human's TRUST in the coverage, and that trust rests on one legible sentence the machinery earns.", "template": "N of {total} declared public nodes examined; M declared optional/inaccessible with reasons; the private methodology generator is not served by design. Completeness predicate: .", "example": "99 of 99 declared public nodes examined; 0 optional, 0 inaccessible; the private methodology repo is not served by design. Completeness predicate: HELD." }, "honest_ceiling": "This protocol gives a reader a route and a checkable completeness predicate. It cannot force an honest read, and it does not judge the report's quality — only whether coverage is complete by the manifest's own node set. Visibility, not immunity." }