Talks

How to turn UI test writing into a sequence of tasks for an AI agent: follow a scenario, identify interface elements, generate a Page Object and check the code. A walkthrough of LangChain and LangGraph, tools, memory and context management, with practical advice on cost and checking the results. The demonstration uses UIAutomator; the approach is also discussed in relation to Espresso.

From imperative UI to an interface as a function of state. This lecture introduces composable functions, recomposition, state and unidirectional data flow, then moves on to building screens: modifiers, layouts, animations, effects and lists. It also covers performance, themes, navigation and View/XML interoperability, followed by questions and answers.

What a dependency is and how the way it is passed affects coupling, testing and architecture. Through dependency graphs, dependency inversion, factories and Service Locator, the lecture builds up to how DI frameworks work. The practical section focuses on Dagger: injection, modules, components, scopes and additional features. It concludes with a look at Hilt’s limitations and an introduction to Yatagan for large projects.

How to reason about architecture without reducing it to pattern names. Ideas from Kant and Hegel, historical development and contradictions provide a way to examine the evolution of MVC, MVVM and MVI. The talk concludes with an example of how this way of thinking can reveal a mismatch between interface state and actions, and suggest the next step in an architecture’s development.

Lessons from automated testing of the mobile Yandex Browser: how the Appium test infrastructure works, why emulators, Docker and Selenoid were needed, and why the team began developing Espresso tests. A comparison of capabilities, speed, infrastructure costs and developer experience, with particular attention to test reliability, CI integration and combining the two tools.

The evolution of functional testing infrastructure at Yandex: from a single Appium server to distributed execution on physical devices and emulators. How to select devices through Selenium Hub, isolate environments in Docker, automate phone connections and start a container for each test with Selenoid. A look at practical failures, scaling and infrastructure optimization.