Scrutexity

Standards Log

Methodology & protocol changes

Institutions evolve transparently. Every change to our observation methodology, pattern registry, or classification rules is logged here with date and scope.

v1.2July 2026
  • Updated claim extraction rules to reduce false positives on comparative claims ("better than", "unlike competitors").
  • Improved evidence classification — split "weak_evidence" into "partial_citation" and "uncited_assertion" for clearer signal.
  • Added rewrite guidance output to the free scanner tier.
  • Refined pattern registry: merged 2 redundant FTC endorsement patterns, added 3 new FDA compounding pharmacy patterns.
  • Hash-chain now includes methodology version in block metadata.
v1.1June 2026
  • Initial pattern registry: 18 FDA/FTC enforcement patterns seeded.
  • Claim health score algorithm deployed (0–100 scale, weighted by evidence status and claim type).
  • Distortion root-cause engine: 8 classification categories at ≥60% confidence threshold.
  • CRT (Claim Record Transparency) receipt format v0.9 specified.
  • Public health endpoint deployed at /api/health.
v1.0May 2026
  • Scrutexity observation methodology published.
  • Claim ingestion pipeline deployed: Playwright crawler → LLM extraction → vector storage.
  • SHA-256 batch hash-chaining implemented for observation integrity.
  • Free scanner (AuditGPT) launched at auditgpt.ai.
  • S-Mark badge specification finalized.