Every week, vendors, analysts, and regulators make specific, testable claims about enterprise agentic AI. Most are never tested again. If you've seen one that deserves review, surface it here.
What passes the significance bar
The claim is specific and testable — a named number, a dated threshold, a quantified prediction.
The source is public — press release, earnings call, keynote, paper, regulator statement. NDA-bound vendor briefings are not archiveable.
The claim is material — procurement-, board-, or policy-relevant at enterprise scale.
A 12–18 month review window is meaningful — either the claim's truth will be clearer by then, or its absence of follow-up is itself the signal.
Full triage methodology at /archive/method/. Submissions that don't pass get a brief reply explaining why.
Disagree with an existing claim's verdict? File a correction instead — corrections move existing records; submissions open new ones.