[00:03]
Google's new Gemini Robotics 2.5 just dropped and robots are doing things that weren't possible even six months ago. A robot tying a ribbon bow, picking up pastries with tongs, sorting car parts inside a real Mercedes-Benz factory, and
[00:17]
hitting a 96% success rate while doing it. That's what we're covering today. And before you think this is just some lab demo, this is live inside actual factories in Hungary and Germany right now. Let's get into it. So first, what
[00:31]
now. Let's get into it. So first, what actually is Gemini Robotics 2.5? Google DeepMind officially announced this on November 10th, 2026. It's their most capable robotics AI model yet. And the thing that makes it
[00:43]
different is what they call a VLA model, vision, language, action. Here's what that means in plain English. Old robots followed a script. You program them to do one thing. They do that one thing forever. They can't adapt. They can't
[00:56]
think. They can't handle anything unexpected. Gemini Robotics 2.5 works differently. It sees through cameras. It understands instructions you give it in plain language, and it converts all of that into physical movement. So instead
[01:09]
of programming a robot step by step, you can say pick up the bracket, put it in the tray, adjust if something's off, and it does it by itself. That's a totally where the numbers get interesting because across more than 200 distinct
[01:23]
tasks they tested, success rates went up 30 to 100% compared to the previous 30 to 100% compared to the previous version. Task completion got 20 to 30% faster. And some tasks that were basically impossible before are now
[01:36]
working consistently. The pastry demo went from 5% success to 100%. The ribbon bow tying flexible material that changes shape constantly went from 0% to 70%.
[01:48]
Folding clothes, precision pipetting in a lab, household tasks, all of it And the reason this is happening isn't magic, it's training. The more scenarios a robot practices, the better it gets. But training real in the real world is
[02:01]
that. Google built something called Gemini 4 Argon. And this is the part most videos won't cover. Training a robot used to mean set up cameras, run the robot, watch it fail, adjust it, run it again
[02:14]
thousands of times. It took forever. Gemini 4 Argon changes that. It can generate thousands of simulated training scenarios automatically. It builds digital twins of real environments from camera footage. It runs practice runs in
[02:27]
simulation, not in the real world. And then it analyzes failures and suggests physical hardware. Think of it like this. Instead of teaching a kid to ride a bike by letting them fall off hundreds of times in real traffic, you give them
[02:41]
a simulator first. They learn faster, they fail safely, and when they get on That's what Argon does for robots. And the result is that Gemini Robotics 2.5 is learning skills faster than any previous version, and it's doing it at
[02:55]
Quick pause before I show you the next part because if you're watching this and automation tools like Gemini to grow your business, the AI Profit Boardroom has a 30-day roadmap built specifically around this. Day one, we cover how to
[03:10]
map every repetitive task in your business that AI can take over. Day two, we get into setting up your first automated workflow. Day three, we go deep on AI content and lead generation systems. And it builds from there every
[03:24]
single day for 30 days. We've got four live coaching calls every week where you bring your actual setup and get real feedback. And right now, there are members inside who are already running AI automation stacks in their
[03:36]
businesses, marketing, outreach, content, customer acquisition, all of in the description or go to aiprofitboardroom.com. So, back to the training problem. Now, let's talk about the factory because this is the most
[03:49]
announcement. Google DeepMind partnered with Apronic, the robotics company behind Apollo 2 Ficus and Mercedes-Benz. They ran a live proof of concept inside real production facilities in KecskemΓ©t, Hungary and Sindelfingen, Germany. The
[04:04]
task was automotive kitting, which means look inside a bin full of mixed parts, them up, place them in the right sequence in a tray. Sounds straightforward. It's not. The parts shift around, the bins aren't always
[04:17]
organized the same way, things go wrong, and the robot has to adjust on the fly. Apollo 2 running on Gemini Robotics 2.5 hit a 96% success rate in a real factory, not a lab, a live production environment. Now, to be precise, this is
[04:32]
a proof of concept. Mercedes hasn't replaced its entire workforce with humanoid robots. That's not what's happening here, but a 96% success rate in a live industrial setting is a genuinely significant result, and it
[04:45]
other thing worth knowing about the safety side, working in a factory means working around people, around obstacles, around moving equipment. Google improved around moving equipment. Google improved Gemini Robotics 2.5's ability to operate
[04:58]
safely. Robots are now 33% better at completing tasks safely around obstacles and 43% better at completing tasks safely around nearby people. They also released something called Asimov Jailbreak. It's an open-source benchmark
[05:13]
designed to test whether a robot can be tricked through disguised instructions into doing something unsafe. The previous baseline for Gemini's embodied reasoning model sat around 82% resistance to these kinds of
[05:26]
manipulations. With Gemini 4 Argon, that went to 99.5%. That matters because as robots get more capable, making sure they can't be manipulated becomes just as important as making sure they can do the task. Now,
[05:41]
you? Here's the honest answer. Right now, Gemini Robotics 2.5 is rolling out to robotics partners, Actronic, Agile Robots, and others in Google's ecosystem. You can't open the Gemini app and control a humanoid robot. That's not
[05:55]
what this is, but what this announcement signals is that the gap between AI in software and AI in the physical world is closing fast. Right now, today, the most version of AI automation is still digital. Marketing automation, customer
[06:09]
support, lead generation, content creation, workflow automation, that's that's where the biggest immediate wins are. But the direction of travel is clear. AI is moving out the screen and into the
[06:21]
physical world, and the businesses that understand this shift earlier are going to have a real edge. The Mercedes-Benz factory hitting 96% in a live proof of concept isn't a lab result. This is production-ready technology moving into
[06:34]
the world, and it's moving fast. Google isn't the only one working on this. competitive spaces in the entire industry. Every major player is pushing you're a business owner, an entrepreneur, someone building
[06:49]
something, the window to get ahead of this is now, not next year, now. Here's a window right now where most business owners haven't figured this out yet. they think it's interesting, and then they go back to doing everything
[07:02]
stay open forever. The businesses getting ahead right now are the ones taking action while this is still early. Inside the AI Profit Boardroom, you get daily tutorials walking you through exactly how to implement AI automation
[07:15]
in your business, specifically around tools like Gemini and the physical AI calls every week so you're never stuck, a full prompt library so you're not starting from scratch, a member map so you can connect with people near you who
[07:28]
are already building this, and a 30-day roadmap so you know exactly what to do description or go to aiprofitboardroom.com. And if you want the full process, the SOPs, and over 100 AI use cases like
[07:41]
this one, join the AI Success Lab. It's our free community. Links are in the comments and the description. You'll get all the video notes from this episode plus access to tens of thousands of members who are actively using AI in
[07:53]
current member count in the description before you watch the next video. It exactly how fast this community is growing. It's free to join, go grab it.