Autonomous agents now write, run and repair tests faster than any team can review them.
But an agent only knows what its environment reports back. Run one against a simulator and it never meets the conditions users actually live with: aged hardware, OEM firmware differences, thermal behaviour, degraded batteries, the OS versions nobody tests on.
Run one against real hardware and it still cannot tell you a device ran flat mid-test, because no test report carries physical state. Either way the failure is invisible in the data. The suite goes green and you find out from a support ticket.
This session is about that blind spot. Nick Metcalfe leads Technical Sales Engineering at Cambrionix and supports customers' test environments daily, from a handful of devices on an engineer's desk to data-centre operations running thousands continuously.
Nick will walk through how one major social platform validates an app used by close to a billion people a day, to show what closing this gap looks like in practice, and why physical validation now has to be as programmable as the agents running ahead of it.
What you will learn:- Where the line actually sits between fast virtual coverage and validation a release - manager will sign off on
- The specific real-world conditions virtual environments cannot reproduce, and which ones do the most damage in production
- Why a test failing on physical hardware has three possible causes rather than two
- What it takes for a device estate to run unattended overnight and return results the team trusts in the morning