Android sideloading isn't going away.
That was the biggest takeaway from my conversation with Sameer Samat, President of the Android Ecosystem at Google. Despite new developer verification and safeguards, Samat confirmed that advanced users will still be able to install any app they want.
We also got into Gemini replacing familiar Google Assistant functions, Google's plans to make Android an “intelligence system,” how AI is changing software engineering, and Rambler, a new voice input feature that edits your thoughts as you speak.
Google's Android figures put the platform at more than three billion active monthly devices worldwide. That's an absurdly big number, and it makes every major change complicated. A feature has to work for someone testing new phones like the Google Pixel 10 Pro or Samsung Galaxy S26 Ultra on launch day and someone whose Android device is the only computer they own.
I sat down with Sameer Samat, President of the Android Ecosystem at Google, at Google’s London office on July 14th, 2026. The full conversation was published on YouTube on July 21, 2026. His Google profile says his role covers Android's global user and developer experience across phones, cars, wearables, TVs, AR and XR, Android Studio, and Google Play.
We got into the frustration around Gemini replacing familiar Google Assistant functions, Google's plan for a hybrid AI system, and Rambler, a new voice input feature that edits your thoughts as you speak. Sameer Samat also explained how Android teams already use AI for engineering work and gave a surprisingly detailed answer about the future of sideloading.
Some of the best details were buried inside longer answers, so I've pulled out the parts that actually matter and added a bit of context around them.
Want to watch the full interview?
I’ve tightened a few answers for clarity, but I haven’t changed what Samat meant. Every direct quote in this article comes from my original interview.
Top takeaways from my interview with Sameer Samat
- Gemini needs a hybrid system. Basic commands stay fast, while harder requests go to AI.
- Android is getting smarter everywhere. Google wants AI built into daily phone use.
- Rambler cleans up speech. It removes filler, corrections, and abandoned thoughts.
- AI is helping Android engineers. Teams already use it for testing and bug triage.
- Computer science still matters. Samat sees it as training in problem-solving.
- Sideloading is staying. Advanced users will still be able to install any app.
- Verification checks identity. Google wants to flag fake developers without approving every app.
The Android settings and apps Sameer Samat uses
Before we got into Android's bigger AI plans, I wanted to know how the person running the ecosystem sets up his own phone.
What does Sameer Samat change first on a new Android phone?
If Gboard isn't already installed, Samat adds it first. He also sets the blue light filter to switch on in the evening.
It's a practical answer, and honestly, I like that. The keyboard affects almost everything you do on a phone, while a nighttime display setting becomes useful every day. If you're using a Google Pixel 10 Pro, my guide to set up your new Google Pixel 10 Pro covers the same kind of early cleanup.
Which non-Google app does he use most?
Samat travels regularly, so he uses Time Shifter to plan when he should sleep before and during a trip. The app builds a schedule around the destination time zone so he can arrive closer to the right routine.
Those questions were a useful warm-up before we got into the real gap here: what Gemini can do in a demo versus what people expect their phone to do every single time.
Android is building a hybrid system for Gemini and basic commands
How will Android keep simple tasks fast and reliable?
I asked about a complaint I see constantly in Android communities. Gemini can handle complicated questions, but basic requests such as setting a timer, calling a contact, or turning off a light can feel less dependable than they did with Google Assistant.
Samat said Google reads that feedback on Reddit and other forums. He also agreed that basic actions need a different standard. Calling your mom or switching on the flashlight should be "faster, better, 100% reliable."
His answer was a hybrid system. Simple commands would go through the fastest and most predictable route available, while complicated requests would use the more capable AI model. Samat described the middle layer as "a router, if you will," with a different engine available on either side.
The tricky part is context. If you ask the phone to call your mom and then make a more complicated follow-up request, the smarter system still needs to know that the call happened. Android has to route the first command cleanly without losing the information Gemini may need a moment later.
That makes far more sense than forcing every request through the same large language model. I've written before about Google Gemini and the impressive things it can do. A clever demo still needs dependable daily behavior, and a phone assistant has to nail the boring commands.
Google wants Android to become an intelligence system
How can Android stand apart when Google supplies AI to other companies?
That question has become even more relevant now that Google is helping power AI on one of Android’s biggest competitors. In January 2026, Apple and Google announced a multi-year collaboration that will see the next generation of Apple Foundation Models built using Google’s Gemini models and cloud technology, including the models behind the new Siri experience.
That means Google isn't just competing with Apple on AI. Some of the underlying technology powering Apple Intelligence is now being built with Google too. So the real differentiator for Android can't simply be access to Gemini. It has to be what Google does with that intelligence when it controls the operating system and can integrate AI throughout the Android experience.
I also asked how Android stays competitive when Google provides AI technology to companies that compete with it elsewhere.
Samat's answer was that Google already works this way. Its cloud business supports many companies, while separate product teams use the same broader technology to build their own experiences.
The more interesting part was how he described Android's direction. Google wants Android to become "an intelligence system." In plain English, AI should show up throughout the way you use the phone instead of sitting inside a chatbot that waits for you to open it.
Samat said consumers aren't interested in "AI for AI's sake." They want daily problems to become easier. That should be obvious, but the industry keeps shipping features that look great in a keynote and disappear once the demo is over.
Android still makes us micromanage plenty of small actions. We decide what we want, break it into steps, then tap, swipe, or type each instruction. The keyboard is the clearest example because every thought becomes words, and every word becomes individual taps.
Google's idea is to let the phone understand more of the intended outcome. That philosophy is also part of what makes switching to Android from iPhone interesting. Android's flexibility matters most when it changes how the phone works around you, rather than adding another menu full of options.
Rambler will edit your spoken thoughts
What is Rambler on Android?
Rambler is one of the clearest examples of this intelligence system doing something genuinely useful. Current voice dictation records what you say, including filler words, repeated phrases, corrections, and mistakes. Samat said Rambler "reinvents the microphone on the keyboard" by working out what you were trying to communicate.
His example was a grocery message. You might ask someone to buy bananas, strawberries, and apples, then remember that you already have too many bananas. Normal dictation keeps the whole messy thought. Rambler drops the abandoned banana request and creates a concise message with the items you still want.
That turns the feature into a lightweight editor rather than a basic transcription tool. You can speak naturally without planning a perfect sentence before opening the microphone.
Google's Gemini Intelligence announcement says Rambler will begin rolling out on select Android devices during summer 2026, with broader availability planned later in the year. That rollout timing comes from Google’s announcement. Samat’s comments in our interview focused on how Rambler works and why Google is building it.
I've seen plenty of paid apps promise cleaner voice writing, so having this built into Android could be genuinely useful. The real test will be how confidently it handles names, technical terms, and changes of mind without rewriting something you meant to keep.
AI-generated code still needs proper engineers
According to Google’s Cloud Next update, 75% of its new code is now AI-generated and approved by engineers, up from 50% in late 2025. Google also says engineers completed one complex migration six times faster with help from AI agents. That gave me an obvious follow-up for Samat.
That gave me an obvious follow-up:
Are software engineers becoming AI prompters?
He pushed back on the label. "I wouldn't call folks who are working with AI tools prompters," he said. I think he's right. System design, architecture, and solid engineering practices still matter because AI can produce code without understanding the larger product decisions around it.
Samat also pointed out that developers have spent years using templates, visual builders, and other tools that reduce manual coding. AI pushes that process much further, but the basic idea has been around for years.
How are Android engineers using AI today?
The most useful example came from Android's London office. Samat described a team that built a bug-triage bot capable of handling critical issues well enough that someone no longer had to sit on pager rotation for that work.
Honestly, this is exactly where AI makes sense. Bug triage is necessary, but very few engineers join Google because they dream about sorting incoming reports overnight. Automating repetitive classification gives them more time for the architecture, debugging, and product work that needs human judgment.
Do programmers still need computer science degrees?
Samat said Google once placed much more weight on technical degrees. When he joined more than 17 years ago, an engineering background was important for engineering and product management roles. Google has since moved toward assessing what a candidate can actually do.
He still sees computer science as valuable. His line was simple: "I think it's a degree in problem solving."
Learning to code and studying computer science give you different things. AI can help someone produce working code, while computer science teaches them to break down a problem, choose an algorithm, judge efficiency, and design a system that remains manageable.
Samat expects programs that focus heavily on teaching one particular language to shift toward broader problem-solving. AI fluency will matter too, but knowing how to ask a tool for code doesn't automatically mean you know what should be built.
Advanced Android users will still be able to sideload apps
Why did Sameer Samat want to discuss sideloading?
At the end of the interview, I asked Samat which question people hadn't been asking him enough. He chose Android openness and sideloading.
That answer matters because sideloading has always been one of Android's clearest differences. It lets you install software outside Google Play, which gives users more choice but also creates an easier route for scams and malicious apps.
Android's scale changes the calculation. Samat pointed to "over three plus billion users around the world," including people who rely on a phone as their only computer. That one device may hold their banking access, private conversations, work files, identity documents, and almost every other sensitive part of their life.
Why is Google adding developer verification?
Google has heard concerns about fraud from users, banks, consumer groups, and other organizations. One common scam involves someone calling a victim, pretending to represent a bank, and sending them a link to install a fake security app.
Under Android's developer verification policy, developers provide identity details and register their apps so Android can connect the software to a real person or organization. Google's current documentation says the system links real-world entities with their Android applications and applies to apps distributed through Google Play and elsewhere.
Samat made an important distinction during our interview, and this is the bit that can easily get lost. Developer verification does not mean Google is reviewing every sideloaded app and deciding if the software is good or bad. It is about confirming that the person distributing the app "is actually who they say they are."
In the fake bank example, Android could warn you that the app doesn't come from the organization it claims to represent. You can still make a decision, but you get much better information before handing a stranger control of your phone.
Will advanced users still be able to install any app?
According to Samat, yes. He said Android will include a mode that an advanced user can deliberately switch on through "a one-time change."
Inside that mode, Samat said users will be able to sideload anything they want, with "no restrictions on it," and the experience will work much as it does today.
That is the clarification many Android users wanted. Google is building a safer default path for someone who may not understand an APK or developer certificate, while preserving a route for people who know exactly what they are installing.
There are still details I'd like to see. Google needs to explain where the setting lives, how strong the warnings will be, and what happens when a scammer pressures someone to enable the advanced mode. The idea sounds reasonable. The friction around that choice will decide how open it feels in practice.
Android’s next chapter is more practical than flashy
My biggest takeaway was simple. Google wants AI to become part of how Android works every day.
The hybrid approach makes sense to me. It could keep timers, calls, lights, and other basic actions reliable while giving Gemini room to handle complicated requests. Rambler could make voice input useful without forcing you to speak like you're dictating a legal document. AI coding tools can take repetitive work away from engineers, but Samat still expects people to understand architecture and problem-solving.
The sideloading plan follows the same logic. Android is trying to protect a huge mainstream audience without locking out advanced users who chose the platform for its openness.
For practical takes on the technology shaping the devices we use every day, subscribe to my newsletter Experiments in Progress.
Every direct quote from Sameer Samat comes from my original interview. I’ve linked any separate rollout details, product information, and Google statistics to the company’s own documentation.
FAQs
Who is Sameer Samat?
Sameer Samat is President of the Android Ecosystem at Google. He oversees Android's global user and developer experience across phones and other device categories, along with developer tools and Google Play.
What did Sameer Samat say about Gemini on Android?
Samat said Android is working on a hybrid system. Simple commands would use a fast and predictable engine, while more complex requests would be routed to a smarter AI system. The systems would also need to share context for follow-up requests.
What is Rambler on Android?
Rambler is an AI voice input feature that turns natural speech into cleaner writing. It can remove filler, corrections, and parts of a request that you change while speaking.
Is Android removing sideloading?
According to Samat, no. He said advanced users will be able to switch into a mode that allows app installs without the standard restrictions. Everyone else will see stronger developer identity checks designed to reduce impersonation and fraud.
Does Sameer Samat think AI will replace software engineers?
No. Samat said architecture, system design, and good engineering practices remain essential. He sees AI as useful for work such as testing and bug triage, while engineers stay responsible for the larger system and product decisions.
