This is your weekly update on the coolest developer news from Google.
This summer the Android folks have been working on 11 Weeks of Android. And this week the talk was all about going beyond the phone to all of the different devices that Android supports. For the watch, we've been investing in the fundamentals of Wear OS, and we focused on things like performance and faster app startup times. And there's a platform release coming later this year.
I personally love having Android on my TV so that I can use apps on the biggest screen in the house. And as a developer, there's a bunch of new resources to help you build your first TV app, including things like Cast Connect and frictionless connections.
This week also had the exciting release of Go version 1.15. It has lots of improvements, including substantial improvements to the linker. And these will work about 20% faster and use about 30% less memory on some architectures. There's also lots of updates and improvements to the core library. So go check them out on the Go blog.
Google Cloud continued the Google Cloud Next OnAir Weekly, and with week five this week it's all about data analytics. There's a whole lot going on in this space, including the launch of BigQuery Omni. And this is a multi cloud analytics solution that lets you query data that's stored across multiple cloud providers, not just Google Cloud.
There's also Data QnA, and this is a natural language interface for analytics. So instead of saying something like, select star from products, order by top sales, you can simply ask, give me the top products. You can learn more about analytics and Google Cloud, including a really cool interactive demo of vehicle fleet management at the blog post.
Google AI released a new model for pose estimation called BlazePose. And this provides pose tracking by employing machine learning to infer 33 landmarks on the body from a single frame. It's uniquely suited for fitness applications. And it's actually small enough to work on mobile phones using only the CPU for inference. If you want to play with it, you can download a version of the model for mobile devices or you can use it in Python. And it's also coming soon to ML Kit. So check it out and download the models from the Google AI blog.
And while we're talking about machine learning, the TensorFlow team announced their new OpenCL-based mobile GPU inference engine for Android. And this can speed up inference times up to 2x. So best of all, using it is actually seamless if you currently use the GPU delegate in your apps. When doing this, the TensorFlow Lite runtime will check on the availability of OpenCL and if it's there, it'll use it. Otherwise, it'll fall back to the existing OpenGL back end.
Exciting news if you're a Chrome OS developer is the launch of ChromeOS.dev. Now, this is a new site that's dedicated to developers, designers, product managers, and business leaders, to help you build the best possible Chrome OS apps. You can see the announcement on the Google Developers Blog.