We tried it. Here's what happened.
In networking, a message only counts as delivered once the other side sends back an ACK. We hold software claims to the same rule.
SYN
You send the claims
List your product and what it can do, one capability per line, with a test account if it needs one.
SYN-ACK
We try each one
An agent opens a real browser, signs in and tries every claim, saving screenshots as it goes.
ACK
A person signs off
Every result is reviewed by a person before it's published. Nothing goes public on the agent's word alone.
The rules
- A capability is confirmed only if the evaluator saw it work.Anything we couldn't check is "not confirmed", never assumed.ACKConfirmed
- We never get around CAPTCHAs or bot protection.If we're blocked, the result says so and the capability isn't confirmed.NAKCouldn't reach it
- A listing nobody has checked says so.Those claims are the founder's word, shown as exactly that, and agents never see them as verified.—Not evaluated
Test accounts
Test credentials are encrypted, used only during that evaluation and deleted when it ends. The agent never sees them: it types placeholders we fill in, and we blur them out of every screenshot.
What the fee pays for
The review, not a result. A capability can fail. Evaluations are free during early access.