Why more than one tester
A page does not fail for “users”. It fails for a kind of user, and the kind you are not is invisible to you. Somebody who lives on their phone and somebody who uses technology because they must give up in different places, for different reasons, on the same screen.
Running twelve segments through one product tells you which audience it already serves and which one it is quietly losing, which is a different question from whether it works.
Where the testers come from
Twelve archetypes, or two hundred records of thirteen dimensions each, from MatrAIx-Persona-8B under an MIT licence. Ethnicity, religion, marital status and gender are dropped at the schema rather than filtered on the way out.
Each tester gets a second-person brief (the real string, not a summary of one) and then picks its own next move, narrates why, and leaves when it has had enough.
A segment is only reported when it converges
A pattern is not filed because one tester hit it. It is filed when every member of a segment hit the same thing and somebody outside that segment did not. That is what turns “somebody struggled” into “you are losing this audience here”, and it is arithmetic rather than a model’s opinion.
Running it
node cli.cjs wander # one tester, free-roaming
node cli.cjs wander --signed-in # past the login
node cli.cjs wander --mission checkout # give them a goal
node cli.cjs focus --panel 12 # twelve segments at once, then the cross-readTwelve testers thinking aloud through a product is twelve transcripts. Start with one
wander to see the shape of the output before you run a full cohort.
What it cannot tell you
Simulated testers are for exploration, not a replacement for real people. They will find the door that does not open. They cannot tell you whether anybody wanted to walk through it, and the product says so wherever this feature is sold.