Talking about Quern on the Evil Tester show

Alan Richardson — Evil Tester — invited me onto Show 033 to talk about Quern. The episode is on YouTube, and his site has the show notes and a full transcript.

Quern is a local, open-source API and MCP server for iOS and Android — real devices, simulators and emulators. It gives an AI agent the things a human tester takes for granted: UI controls, screenshots, logs, and network traffic. Mobile testing has always front-loaded a lot of setup before you can observe anything at all, but collapsing that is just the start of what Quern is for. Through Quern, your agent can now operate and observe your device in ways that were impossible before, letting you be in the passenger seat, giving directions, while your AI agent drives.

The framing we kept returning to is that this is pairing, not delegation. The agent is driving, but it drives in the way that a rally car driver does - by listening to the navigator in the passenger seat. It widens what you can observe — it can watch more things at once than you can — but the tester stays in the navigator seat. That distinction matters more than it sounds, because the failure mode of handing an agent the wheel with no directions is that you both can get lost.

We also got into turning exploratory sessions into repeatable scripts, where current models fall down, and why mobile testing might be heading somewhere genuinely different rather than just faster.

Thanks to Alan for the conversation — and for asking better questions than I was ready for.