Proof, not promises
Don't trust the number. Reproduce it.
Most research asks for your faith. We hand you the evidence and the method. Run it again on your own machine and reach the same verdict.
Recall on our internal planted-defect corpus — benchmark fixtures we authored ourselves, with the honest caveats published alongside, not hidden.
Self-graded against defects we planted, not independent third-party validation.
Bytes that leave the machine. Unplug the network and every result still reproduces.
Same input, same verdict, every run. The deterministic core never drifts.
How we prove it
Three ways to check our work.
Reproduce
Run it offline
Point AffectEQ at the same page on a disconnected machine. The deterministic audit returns an identical score.
Trace
Follow every finding
Each issue links back to the exact witness that raised it. No black-box verdicts, no unexplained numbers.
Disclose
See what we filter
Where the AI pass discards its own low-confidence findings, we say so, and we show you the rate.