Android is transcending its role as a mere operating system to become a unified cognitive architecture where apps serve as modular functions for a central intelligence. This shift marks the definitive end of the siloed application era in favor of a seamless, intent-driven ecosystem.
深掘り
前提条件
- データがありません。
次のステップ
- データがありません。
深掘り
Connect to the intelligence system追加:
[music] [music] [music] [music] [music] [music] [music] [music] [music] >> We're creating a platform that delivers new, helpful, AI-powered experience that actually anticipate users' needs. Now, for you, this means not only new discovery surfaces for more and deeper app engagement, but users are actually getting better and safer app experiences. And we're doing this at every level of the stack. This is what we mean when we say that Android is evolving from an operating system to an intelligence system.
Now, your apps will be at the center of these changes in different ways. First, let's talk about how AI can enhance the ways users interact with your app with proactive help, leveraging the platform's deep understanding of native UI built with Kotlin and Compose.
An example of this is our next-generation autofill, which works to get your app higher form completion rates, all using the same hints you're currently providing autofill.
We can also help protect your users today on Pixel with using AI to identify scam patterns and notifications from popular messaging apps, warning users before they respond to a potential threat.
Now, Magic Compose was introduced last year on Pixel 10, and it surfaces helpful information and actions within apps. And this is a new place for users to engage with and discover app content.
And as you can see here, it's expanding to more apps, such as Snapchat.
Now, improving your apps with proactive AI features is awesome, but we're entering the agentic era, where agents and assistants complete tasks on behalf of users, and your app can participate.
I mean, after all, like, Android has always been designed to have multiple apps work together to share a task, such as composing an email or with a photo.
But now, agents such as Gemini can act as an orchestrator across multiple apps.
Now, to this end, Gemini can analyze app screens and navigate through tasks, stopping for user input when it needs clarification at sensitive areas like transactions, with no code needed from you in almost all cases. The user can do other tasks while Android actually runs the app in the background on a virtual display. And we'll be rolling this experience out more widely in the future.
But, you can take control of the capabilities your app exposes by integrating with a new API, app functions.
Now, app functions is an Android platform API and a Jetpack library that simplifies building Android MCP integrations. It empowers your app to behave like on-device MCP servers, contributing functions that act as tools for use by client agents and assistants like Gemini. Now, app functions integration with Gemini is currently in a private preview with trusted testers, but you can begin preparing your apps now using app functions.
Now, if you built your app with, you know, clearly separated data presentation layers following our recommendations, adding app adding app functions can be really really straightforward. So, let me show you what that looks like.
All right.
So, here we are in on our laptop.
Let me actually show you a little bit of what we've done.
Uh we've actually got a skill that can actually add app functions automatically to your project.
And it it works even better if you actually have architecture app well, but it can work in all cases, really.
And the key thing is that we're going to do this today with the Android the Now in Android news aggregation app. So, let me quickly just show you if you haven't seen this.
Um this is what that looks like. Very nice, you see? We've got composed first um as part of the news that we're aggregating. Now, what the agent has actually done though is pretty cool. Um so, you can see we've just given it a very very simple prompt.
Create app functions to provide access to topics and news articles, enable users to follow and things like that. It activates the skill. We we've installed that in the project, and it's going to do a lot of things for you. It's going to upgrade things. Let's scroll to the bottom and look at all that it's done.
So, again, we've got we've updated the SDK, we've added dependencies, we've created metadata, and more importantly, we've actually implemented four useful app functions: to list topics, list news articles, follow topics, and and bookmark things.
And this is the actual code that was generated. Like as you can see here, we're really literally running all this from Android Studio, and I have not made any changes to the project. This this actually worked the first time, which is always nice. Um so then you can see it's created this app function for list topics. And what's really great about the way that Jetpack library works is that you can see that this is pretty normal. We've just We've just annotated it with app functions. Um it's a normal suspend function that runs with a context. You can see here that it's actually returning a fairly reasonable data model.
And uh hold on, let me do that one more time. There. A very reasonable data model. And um and also we're documenting it using KDoc. And the library is just going to make sure that that gets all the way to the agent that's using it.
And that's the same thing that makes me understand it, also makes our agents understand it. And you can And you can see we can list topics and news articles. We can follow topics and we can create bookmarks. Now let's But how do you test this? Well, we've actually gone and created an agent that can do that for you. So let's go to our running device and let's switch over to our test agent.
All right. So first of all, we want to do things deterministically. So let's take a look at what functions we're exporting inside of Now in Android. So we've got list topics. We can run those functions. And you see we get some lovely JSON. And you notice ID number three is compose. All right. We'll use that when we go to list news articles.
We can also add parameters here right in the test agent and run them.
And again, beautifully, this is actually happening by the way. This is not like canned or anything like that.
All right. Let's Let's talk about what the next step next thing we want to be able to do is actually run against a live instance of Gemini. So we've hooked it up to a regular cloud instance of Gemini. We can do things like, "Hey, show me the article on compose UI UI first." See, as you know, I just so you know, it's not really canned cuz I would never make a typo on a canned demo. All right. Let's demo gods. Let's Let's do this.
All right. So it's connecting to the Now in Android app.
And Gemini actually has to do a bit of work here because we don't you notice we didn't have a great API for actually searching this. So it's actually pulling the topics, pulling all the articles, and then giving you the article that you selected. And And Gemini is doing all that work. But we can take it one further. Let's Thank you.
>> [applause] >> We can also say send the Compose UI first article to Nick Butcher.
Ah, just to Nick. All right.
So now what we've done is we actually have a second app that uses app functions that actually is a chat app.
We're connected that. We're going to find out who Nick is, first of all. Uh and then we're going to go and send that article to him. It's actually recalled our app. And so now if we go over to our chat application.
Eventually, come on. I can I can use this in an emulator, I promise. All right. There we go. This is the article, Android UI development is Compose first.
So that is again, the agent working together to with with two different apps using app functions. You can imagine all the cool things this can do. All right, back to our slides.
So check out our tools and you can experiment with app functions today. You can join our early access program for the chance to be among the first apps to deploy app functions to production.
関連おすすめ
OpenHuman VS Hermes AI: Who Wins?
JulianGoldieSEO
285 views•2026-05-29
Long-Running Agents — Build an Agent That Never Forgets with Google ADK
suryakunju
142 views•2026-05-30
This computer is made from real human brain cells. And you can buy it.
Talktmsmedia
3K views•2026-05-28
BREAKING: Microsoft’s New Image Generating Model Beat Out GPT 1.5 and Nano Banana 2
aimmediahouse
122 views•2026-06-03
I Made the Same Anime Fight Scene in Every AI Video Generator
NobleGooseAnime
295 views•2026-05-30
Nvidia Bets Big On AI PCs | New Chip To Power Windows Laptops | Technology | AI Updates | N18S
cnnnews18
3K views•2026-06-01
I Tested NEW Opus 4.8 on Four Projects (Updated LLM Leaderboard)
AICodingDaily
298 views•2026-05-29
3D Platformer Update - NO CAPES
SolarLune
294 views•2026-05-30











