Skip to content

Evidence Ledger

The evidence ledger is the archive’s source-traceability spine. It collects every processed source record, concept occurrence, glossary entry, equation candidate, figure candidate, quote candidate, and promoted original scan crop into one machine-readable index.

It is not a book of finished claims. It is a custody and confidence map that tells researchers what exists, where it came from, and how much trust the archive currently assigns to it.

Open the raw JSON ledger on GitHub

Record TypeCountMeaning
Sources11Seeded books, lecture collections, and raw/intake records.
Concept records206OCR/PDF-text occurrence indexes for Steinmetz concepts across seeded sources.
Glossary records134Older or technical terms that require source-wording review.
Equation records2,645OCR/PDF-text-derived equation candidates, mostly requiring scan triage.
Figure records326OCR/PDF-text-derived figure candidates.
Promoted original figures15Scan-crop assets with manifests and checksums.
Quote candidates8Hidden-gem OCR/PDF-text candidates awaiting scan verification.
ConfidenceCountReader Rule
source-custody-record11Use as source inventory, not as interpretive proof.
ocr-occurrence-index312Useful for discovery; do not quote as exact wording.
source-located-ocr-candidate2,979Has source/line location, but the OCR and formula/text still need scan review.
source-located-candidate28Has a source reference, but requires review before canonical status.
scan-crop-promoted15Derived from a rendered source page and backed by a manifest and checksum.

The original project goal was to avoid hallucination while still allowing ambitious interpretation. The ledger is the mechanism that makes that possible:

  • A Steinmetz source claim must point to a source record.
  • A mathematical claim must point to an equation record or derivation page.
  • A diagram claim must point to a figure record, scan crop, or redraw manifest.
  • An ether-field reading must remain interpretive unless the ledger contains an explicit Steinmetz passage.
  • A Tesla-era comparison must be added as comparison evidence, not merged into Steinmetz’s own voice.
  1. Candidate appears in OCR, figure extraction, glossary extraction, or concept tagging.
  2. Candidate enters the evidence ledger with confidence and status.
  3. Researcher checks the scan, page, formula typography, and surrounding context.
  4. Record is promoted into a concept, equation, diagram, glossary, hidden-gem, or comparison page.
  5. The public page links back to the source page and keeps interpretation layers separate.