Lite is free. Pro adds this and the rest of the quality layer.
It says done.It is not.
An AI reports success in the same calm voice whether it ran the check or skipped it. Here is how to tell the difference.
Pro
// the problem
What goes wrong
Done is a sentence, not a resultA model describes what it intended to do. If the test never ran, the report still reads as finished. The words carry no evidence.
The gap shows up laterThe button that does nothing, the page that breaks on a phone, the migration that never ran. You discover it when a user does.
// what adams does
What Adams does
Names the proof firstPro: before a claim of done, the verify step writes the check that would fail if the claim were false, then runs it.
Reads the real outputPro: a skipped check or a warning nobody read counts as not run. It is never reported as a pass.
Reviews with a strangerPro: a separate reviewer, not the author, looks at the change and ranks what is wrong.
// worked example
An example
A claim with its proof
Illustrative session.
~/project~/your-project · example
/adams:verify the signup page works
Claim: signup creates a user and shows a thank you
Proof: run the signup test, open the page at 375px
Fail: thank you message missing on phone width
Fix and run the same proof again
// faq
Questions
What is a good proof?
The narrowest real check: a test that covers it, a build, or a page opened at the widths that matter. A check that cannot fail is not proof.
Can I do this without Adams?
Yes. Ask for the command that proves it, run it, and read the output yourself. The guide on reviewing before deploy covers it.
Is this Lite or Pro?
Verify and review are Pro.
// related