Join hosts Tor and Romain as they sit down with Søren Gjesse, Chris Craik, and Shai Barack.

This week, the team dives into R8, the whole-program optimizer that handles the "last mile" of your code before it runs on a device. Learn why implementing R8 is effectively like giving your users a "free hardware upgrade" by making their current devices perform like models that are years newer.

Enable app optimization → https://goo.gle/R8-Optimizer

Chapters:
0:00 - Intro
0:23 - What is R8?
1:16 - R8 vs D8
1:40 - Tree Shaking & Dead Code
4:14 - Handling Reflection
10:10 - Performance Co-Optimization
13:03 - Kotlin-Specific Gains
16:22 - Real-World Impact
41:50 - Jetpack Compose & R8
50:51 - Keep annotations
1:03:49 - The "Free Hardware Upgrade"

Join hosts Tor, Chet, and Romain as they sit down with Diego Perez (Android Studio) and Patrick Fuentes (Developer Relations) to explore the new frontiers of Android XR.

This episode of Android Developers Backstage breaks down the latest announcements and Developer Preview 3, offering a technical look at how developers can start building for spatial environments today.

Resources:
Hello Android XR sample → https://goo.gle/3Nbnrre
Build for AI Glasses with the Android XR SDK Developer Preview 3 → https://goo.gle/49lUjGr

Chapters:
0:00 - Intro
4:42 - Latest announcements
5:46 - AI Glasses
8:28 -Developer preview 3
16:13 - Spatialization
18:29 - Tools team perspective
24:07 - Using the XR Emulator
29:00 - What is an Orbiter?
37:33 - What hardware is currently available
38:39 - Final thoughts

In this episode Tor and Chet chat with Raluca Saucic and Eugene Yakavets about the new "Bring Your Own Model" feature in Android Studio which allows using other models than Gemini for agent mode, including models from Anthropic, OpenAI, and more. We also discuss using local models via Ollama and LM Studio.

Resources:
Use a remote model → https://goo.gle/43N2Z5E

Use a local model → https://goo.gle/3Xij27V

Chapters:
0:00 - Intro
0:49 - Bring your own model
4:13 - Bring your own key
7:24 - Anthropic model
13:17- How often do you need to remind the model of the context?
17:40 - Thinking vs planning
20:25 - What happens when the API needs to give multiple answers to a complex question?
24:14 - Supported models
31:18 - Are there any barriers when using a different model?
34:07 - Support for code completion
38:16 - GitHub's relationship with agents
44:35 - Using agents for development
56:36 - Studiobot
57:56 - Wrap up