AI & ML interests

None defined yet.

Recent Activity

genstrataeh  updated a dataset about 13 hours ago
Gauntletscore/llm-aggrefact-predictions
genstrataeh  published a dataset 3 days ago
Gauntletscore/llm-aggrefact-predictions
genstrataeh  updated a Space 3 days ago
Gauntletscore/README
View all activity

Organization Card

GauntletScore

AI always sounds sure. Now you can be.

GauntletScore checks the claims in a document against real sources and shows you which ones hold up before you act on it.

How it works

  • Checked against the record: clinical trial registries, published research, court records, company filings and registries. Not how sure the sentence sounds.
  • Independent of the author: it checks the same way whether a person or an AI wrote the document. AI models pull out the claims and read the sources. Math computes the score.
  • Honest about limits: a claim it can't settle is reported as unsettled, not false.
  • Auditable: every verdict records which source the claim was checked against.

Evidence engine (in pilot)

Tested on 2,192 statements across 71 documents it had never seen (Run 28, October 4, 2026):

  • 98.8% right when it decides
  • 0.4% confidently wrong (9 of 2,192)
  • 0 of 415 true figures from company filings wrongly called false

Ways to use it

  • Paste or upload a document at gauntletscore.com
  • API for software companies and agent makers
  • Chrome extension, coming soon

Who it's for

Legal, insurance and finance teams checking documents they received, and anyone checking their own AI writing before it goes out.

Patent pending.

gauntletscore.com · Free to start, 200 claims a month, no card needed.

models 0

None public yet