Gemini Now CONTROLS Your Phone and ORDERS Your Food For You!

This one dropped today alongside the Galaxy S26 announcement, and it's a bigger deal than the headline makes it sound. Google just announced that Gemini can now operate your phone on your behalf. Long press the button, tell it to book you a ride or reorder your last Door Dash meal, and then put your phone down. Gemini goes and does it, navigating the app, filling in the details, completing the steps while you're doing something else entirely. It shows up as a notification you can monitor, jump into, or stop at any time. They're calling it automation and is launching as a beta on the Pixel 10 and Galaxy S26 starting March 11th, US and Korea only. It's also limited to a handful of rid share food and grocery apps to start. The technical approach is worth understanding. Gemini doesn't give free reign of your phone. It runs the necessary app in a secure virtual window and what happens in that window is processed in the cloud. So it's sandbox. It can't see the rest of your device. You can actually watch it scroll, tap, and type in real time if you want or just let it run in the background. Now sandboxing matters because the obvious question with something like this is, do I really want an AI clicking around inside my apps? Google clearly anticipated that push back. The framing is all about control and transparency. You started it, you can stop it, you can see everything it's doing. But here's the thing. The apps it starts with are intentionally boring. Uber, Door Dash, grocery carts, that's not an accident. These are tasks people do constantly. They're annoying and there's very little that can go wrong. It's the safest possible on-ramp to something much bigger. Analysts are already saying this is a sign of where smartphones are headed. a world where users do far less tapping and swiping between apps and AI agents do the heavy lifting. And once people get comfortable with an AI booking their Uber, the logical next step is an AI that can handle more complex multi-app workflows. Schedule a meeting, then book travel around it, then add it to your calendar and send a confirmation email. That's all the same category of task, just more steps. Apple is working on the same thing. By the way, an overhauled Siri with deeper app control has been in development for a while, though it has been delayed. So, this isn't just Google. It's a race to own the layer between you and your phone. What Google has right now is limited, but the pattern is familiar. Start with something small, get people used to it, expand the scope. The jump from Gemini orders your dinner to Gemini manages your morning is smaller than it

seems. So, Miniax just dropped something called Max Claw. And the reason is interesting isn't the name, it's what it actually represents for where AI agents are heading nowadays. We had Kimmy drop Kimmy Claw and now it looks like Miniax is also following the same wavelength. Maxclaw is basically three things combined into one product. You got Open Claw, an open- source AI agent framework that went from 0 to 100,000 GitHub stars in about 2 months. You got M2. 5, Miniax's newest model, and you got Miniax agent, their hosted platform. If you put them together, you get an always AI agent. No server to spin up, no API keys to configure, just running 24/7. Here's the thing about OpenClaw. Before this, it was genuinely powerful. You could connect it to Telegram, WhatsApp, Slack, Discord, and have it running tasks autonomously while you sleep. But you had to host it yourself. You had to manage the infrastructure. For developers, that was fine. But for everyone else, this was an obstacle. Max Claw knocks that wall down. Miniax is essentially saying, "We'll run it for you on our infrastructure 24/7, and we won't charge you extra API fees on top of it. " The reason they can pull this off economically comes down to M2. 5, which is a 230 billion parameter model, but it uses a mixture of expert architecture, meaning it only activates about 10 billion parameters per token that keeps inference fast and cheap. output costs between$1 and $2 per million tokens, which is roughly a tenth of what you would pay for compared to using a Frontier model. And the benchmark numbers are legit. 80% on software engineering bench verified puts it in the same conversation as Cloud Opus at a fraction of the price. Zoom out and there's a pattern here we're going to keep seeing. An open-source project like OpenClaw gets traction because developers love the control and people love it. Then someone productizes it like Kimmy and like Minia Max adds hosting removes friction makes it accessible to people who don't want to touch a server and OpenClaw stayed open and Maxclaw is just the managed version. Both can coexist. The more interesting question is what a world looks like where millions of people have a persistent AI agent running in their WhatsApp or Telegram handling tasks around the clock. We're not there yet, but products like this are on the ramp. Miniax is a Chinese AI company that doesn't get nearly enough coverage in Western tech media. M2. 5 is legitimately competitive at the frontier and Maxclaw is a smart distribution move. This is a story worth keeping an eye on. So the claim by Miniax is that to set up your Maxclaw, your 24/7 personal assistant is supposed to be very easy and straightforward. So I'm going to see the process myself. So I'll click on start now. And then it's asking me to choose an expert. I can you know Miniaax agent seamlessly integrates across popular expert deployed on openclaw. So I can choose if I want a default one a visual lab one a trend intel one which is a class platform social media trend monitor multi- aent trading if you are doing some trading work. So it's pulling up some popular experts. So for my reasons, I will say I want something as trend intel maybe or a tra or a topic tracker, but I'll do trend intel for now. And then I'll press I'm ready. So now it's deploying my agent. So let's see how long this takes. So looks like it only took about 30 seconds at max to set up my agent, but it says meet Maxclaw chat right here. No setup needed. Just type your first message below or take Maxclaw to your favorite apps and you can connect it to things like Telegram, even Slack to sync your experience across platforms. Once started, the possibilities are endless. So yeah, we can obviously schedule Chrome jobs or also do some complex automations. And it's all going to be powered by the Miniax M2. 5 model. So we can give it a try here. Like I've shown in my Kimmy Claw video how to set it up on Telegram. So I'm guessing it's a similar process. But let's just say if we wanted to uh chat, we can just chat here or we can transition it to Slack or everything like that. The Kimmy Claw one, if I remember correctly, when I was setting it up, there was only Telegram availability, but based on the announcement by Miniax, you can also do WhatsApp, Slack, Discord with Maxclaw and Telegram. So, you have that availability. So, this is really sick. We can give it a personalization, give everything like that. Let me know if you guys want me to do a video where I actually set this up properly, my Maxclaw, and do a full deep dive on it. Apple just announced that they're

expanding their Houston factory specifically to boost Mac Mini production. The campus footprint doubles. They're adding a 20,000 ft training center, and it marks the first time that the Mac Mini will be made in America. Now, why the Mac Mini specifically? Earlier this year, something unexpected happened. The Mac Mini started flying off shelves. Not because of a new product launch or a marketing push, but because of OpenClaw, people started buying high memory Mac minis specifically to run it locally 247 as a personal AI agent. The 24 GB and 32 GB configurations went on back order for weeks. Resellers started listing them above retail on eBay. Tim Cook publicly acknowledged Apple was working to catch up on memory supply. So, this factory expansion is Apple responding to that. a real demand from a direction nobody at Apple predicted. So, OpenCloud did not only push a lot of AI models to start creating their own open cloud versions like we saw with Kimmy and we're seeing subtly with cloud as well as they're dropping automation task and cloud remote all these features. But it's also forcing the infrastructure layer to change by forcing Apple to ramp up Mac Mini production. That's it for today's

video. Make sure to subscribe to the channel, follow us on Twitter, follow the world of AI, and don't forget to subscribe to our newsletter. We post constantly and you don't want to miss this.