From corpus to proof PT-BR

Visual course · 6 lessons · PT-BR + EN

From corpus to proof

Six operational lessons from a full-corpus internalization: truth with receipts, coverage before interpretation, gates that admit no bypass, and an honest mirror of your own sessions.

Sanitized retelling: every example is synthetic or uses patterns from a public seed article. No host paths, session content, secrets, or private data appear in this course — and no page makes a network call.

From corpus to proofFour blocks connect inventory, receipt, gate, and truth.Total inventoryReceipt per itemAdversarial gateTruth with a class
2,775

primary items, 100% with a final state

535,326

verified receipts, 0 problems

1,087

classified claims, 2 in CONFLICT

Local progress

0 of 6 lessons marked complete in this browser.

Track

  1. 1The Truth Contract
    Eight truth classes and the golden rule: repetition, popularity, and author confidence never promote a claim.    15 min
  2. 2Coverage before interpretation
    Inventory first: a final state per item, frozen manifests, and named drift — never silent adjustment.    14 min
  3. 3Proof gates & total coverage
    A gate that doesn't cover everything gets bypassed: seven rounds of attack, minimal repair, re-attack.    18 min
  4. 4Visual inspection discipline
    Models hallucinate confidently on tiny images: the permanent ≤128px rule.    16 min
  5. 5The receipt graph
    Every derived value points to a receipt: snapshot hash, offset pairs, and content-addressed ids.    17 min
  6. 6The mirror
    Mining your own sessions with the 3-dated-occurrence rule: conclusion inflation, repetition tax, and the hard truth.    19 min

Where the lessons come from

From a real internalization mission: 2,775 primary items, 5,386 media paths, 45 sampled sessions, and an independent audit. The run's canonical artifacts (final-report.md, LOOP-LOG.md, claims.jsonl, mirror.md) live in the run directory, outside this self-contained course.

Core rule

Exist → cover → prove → only then interpret. No arrow may be presumed, and every partiality stays visible as partiality.

How to study

Each lesson has plain language, a collapsible technical layer, a deterministic simulation, and a checkpoint. Mark your progress — it stays in this browser only.