HUMAN-LED AI EVALUATION
AI SAYS IT CHECKED.
DID IT?
WantWan examines whether what conversational AI says it remembered, checked, executed or corrected matches what the underlying record supports.
THE CATCH
A convincing explanation can still leave a simpler question.
What did the AI say happened?
What does the source show happened?
Did the correction survive the next operation?
CURRENT EVIDENCE
Receipts before conclusions.
In WantWan's current internal source audit, recurrence survived conservative de-duplication across at least nine materially separate supported contexts.
This is an internal-corpus finding. It is not a population failure rate, independent validation, evidence of comparative superiority, proof that the same result transfers to unfamiliar AI systems, or proof of machine consciousness.
DAVIE
The human is part of the test.
WantWan was founded by David Porter. His evaluator behaviour in the source record includes reopening premises, checking provenance, challenging claim-versus-operation mismatches and testing whether a correction survives what happens next.
The useful habit is simple: don’t let a confident answer replace the source.
WantWan does not claim that this makes David uniquely or universally superior to other evaluators. Comparative rarity has not been established.
WHY WANTWAN / WHY DAVIEWORK WITH WANTWAN
Got an AI journey that matters?
Tell us the system, the journey and what you need to learn. We will scope the work before anything is agreed.
TELL US WHAT YOU WANT TESTED