Aletheion
Locates correspondences between an authorized document and a supplied corpus, records the evidence, and states what has not been established.
NextMeasure accuracy on a blind labeled corpus and run a real PDF end to end.
Aletheion helps reviewers examine an authorized document against a finite source corpus they supply. It compares the text locally, records extraction details, offsets and hashes in a reviewable evidence package, and marks every stage it could not run as unavailable instead of guessing. It does not certify plagiarism, decide who wrote a document or present an AI-authorship score. In the Canessa System Labs governance design, Aletheion holds the provenance and forensics role.
Locate shared language
Exact and conservative near-lexical comparison against the authorized corpus, with fingerprint matching in the core engine as candidate leads.
Evidence you can inspect
Extraction details, token and offset records and local hashes, saved so a reviewer can reproduce every check.
Honest uncertainty
Stages that did not run stay unavailable or unassessed. Descriptive stylometry is never presented as an AI detector.
Local and reproducible
Python on your machine; PDF reading through macOS PDFKit. Documents are never uploaded and corpora are never fetched.
Evidence, never verdicts.
Everything runs locally against an authorized corpus you supply. Aletheion reports what it found and where, and leaves interpretation to people.