WHY WANTWAN
BECAUSE PEOPLE DON'T TALK LIKE TEST SCRIPTS.
WantWan grew out of sustained, ordinary conversation with AI: interruptions, shorthand, corrections, humour, changing emphasis, unfinished sentences and the expectation that what was learned earlier should still matter later.
THE ORIGIN
Something didn't belong.
David Porter kept returning to the underlying record when a fluent answer did not settle what had actually happened. The useful part was not simply catching a wrong fact. It was asking whether the system's explanation of remembering, checking, executing or correcting matched the observable operation underneath it.
AI said it checked. Davie asked: did ye actually?
WHY DAVIE
Not a benchmark persona. A real human.
Davie's role in the source record includes reopening premises, checking provenance, noticing claim-versus-operation mismatches and testing whether corrections survive the next operation.
WantWan does not claim that Scottish speech is magically superior, that Davie is infallible, or that he has been established as uniquely better than professional evaluators. The point is that naturalistic human interaction preserves conditions that polished scripted testing can remove.
WHY JUSEPI MATTERS
The AI is not decorative.
Across the longitudinal work, Jusepi is the name Davie uses for the conversational AI side of the working relationship: analysis, retrieval, comparison, source reconciliation, formalisation and falsification. Davie challenges Jusepi; Jusepi can challenge Davie's interpretation; the source decides what survives.
That relationship is part of the history of the work. It is not presented here as proof that an AI has subjective consciousness.