Synthetic user populations that browse a real web app and turn what they do into findings such as rage clicks, confusion and pages nobody can find.
Understudy is in development, and its detection study is complete.
Understudy runs scale tests of a product with synthetic user populations in place of human testers. AI personas browse a real web app the way a test panel would, and what they do becomes findings, rage clicks, confusion, funnels that break, pages nobody can find. The aim is signal before launch without recruiting hundreds of people, acceptance, conversion and market response simulated at population scale, with the confidence of those findings against real human behavior measured and published per claim type. A tiered decision architecture keeps simulated sessions cheap enough to run in the thousands, and the detection study below drew the first trust boundary for what its findings can back.
Stress test
47.4% of sessions that reached the fully dead chip were caught, in activation mode, in the first population.
Engine spend was $149.51, measured, across the probes, the pilot, the freeze rehearsal, the detection run and the resampling run.