As AI shifts from passive assistants to on-device autonomous agents, mobile software is undergoing its most significant platform transition in decades. Android is evolving from a visual app launcher into an intelligent runtime where system-level agents orchestrate multi-step workflows across applications. In this opening keynote panel followed by open QnA, engineering and developer advocacy leaders from Google unpack how the platform is adapting to an agent-first world. The panel examines the shift toward AppFunctions and on-device tool calling, privacy boundaries for autonomous execution, and why developers must start treating app capabilities as structured APIs alongside visual screens. Developers will walk away with an insider look at Android’s platform roadmap and practical guidance on preparing production codebases for an agent-driven ecosystem.

Stop fighting brittle UI locators and flaky IdlingResource boilerplate. This introductory session explores Journeys for Android Studio, showing how Gemini’s vision and reasoning capabilities execute functional test steps written in plain English.

When off-the-shelf system models aren’t tailored enough, bringing custom open-weight models (like Gemma) to the device is the answer. This talk dives into optimizing, quantizing, and executing custom open models directly on mobile GPUs and NPUs.