Boston Dynamics Reveals 2026 Atlas Endgame (AI NEWS)

The core difference is that you don’t want to program the robot explicitly, instead you want to teach the robot how to do a certain task. And as the robot is doing it, you want to tell it when it has failed, how to correct four mistakes, how to improve its performance over time. So there’s a core workflow there, something that we refer to as post-training. And a certain number of times, it could take in the order of a day, for example, of demonstration of an expert demonstrator, to tell the robot how to do that task and how to correct for certain errors that might happen during the execution.

And accumulate all that knowledge, now not in the form of a program, an algorithm, but rather in the form of knowledge ingrained in a neural network. So the most interesting part of this newest strategy is what we call pre-training, which basically accounts for a system that makes a good initial guess as to what should be that policy. So that once you go to factory and do the first deployment, the behavior is already somewhat good, and it takes a small amount of time to improve it. Next, they talk about the future of dexterity using this robot brain.

For the phase of the program where Atlas is, we wanted Atlas to be able to grasp many things. And building a gripper with three fingers, we deemed sufficient. And we’ve proven that it is sufficient. It’s a gripper that you can grasp almost anything that we throw at it. If you add more fingers, if you add more complexity, it always ends up, you pay the price in terms of unreliability, you pay the price in terms of like lower speed of development. We are definitely working on more complex grippers that we’ll get to see in sometime in the future, and they might have more fingers if we think that they’re necessary.

Our next milestone is Dexors manipulation. And we are building grippers that have the degree of robustness and strength that our current grippers have, but that they will also be able to use tools, they will be able to do assembly, and they will be able to grasp small objects and handle them. But when will this robot be ready for the consumer market so that anyone can buy it? Here’s what Boston Dynamics says about the timeline. We are doing work to make sure that humanoids are standing in buildings, we won’t build a good product unless we do that.

And so we’ve already started going onsite with Hyundai, which is our first foray into doing onsite engagements. Now, Hyundai is a great partner to start with, because we really need to be working with people in the early days that understand that this is not a hardened product yet. And Hyundai is really on that journey of kind of eating, they call it eating your own dog food when you’re building a product, and working with us as we fail, as we succeed, to build something that really works for operators in their facilities that we can eventually leave on site, and we can call a hardened product that you can call me and buy.

We’re not quite there yet. And we will be getting there over the next year or two. Over the next few years, as we get the product more reliable, that’s when we’ll actually start selling Atlas. And today, you can’t just call a salesperson and buy an Atlas, you have to go through me and Alberto first. And here’s where it gets really interesting. Can you actually build humanoids that can provide that degree of generality? The problem is really hard, right? Like, if you look at an assembly line, the real problem that we’re talking about is someone fishing with its right hand into a bowl full of bolts, picking one, going with the other hand to pick up a bolt driver, inserting the bolt in the bolt driver and screw it on the underside of the roof of a car, maybe while bracing with the other hand to maintain balance.

It is very complex, but that is our lighthouse. That is the kind of task that we are aiming for Atlas to do. There’s fundamentally no reason why it shouldn’t be possible. The hardware is capable of it. We know that. And the techniques we’re using today should be able to scale to that level of complexity. So we’re convinced that it is possible to get that generality. Plus, Boston Dynamics also discussed just how far behind Atlas is from spot right now, to give you a better idea of how fast they’re progressing towards a general release.

Because of the experience we have from spot and stretch, we kind of have an understanding of what is the process of maturing these products, building the hardware, maturing the software, maturing the customer relationships and building trust with them. And in a way, I think of spot is leading the charge. Spot is at that sweet spot right now, where at the sweet spot of the hockey stick, where we have thousands of robots out there in the hands of customers, many customers that are scaling. And you can think of stretch as being two years behind that and Atlas being two years behind stretch.

So I think I’m comfortable saying that within four to five years, we should be where spot is today. In fact, Boston Dynamics even describes how you’ll get Atlas to learn a brand new task in the future. And this brings up an interesting question of whether or not this would qualify as robot AGI. Can you get robots to learn what it means to interact with the physical world with dexterity and common sense by looking at videos of people repairing bicycles in their backyard, or looking at videos of people constructing things. So that’s the extreme of it.

And finally, generative AI just experienced a breakthrough with Google having released its Gemini 3 flash, expanding its latest model family with what the company calls frontier intelligence built for speed. And the model combines Gemini 3 pro level reasoning capabilities with significantly faster processing and lower costs. So getting into the benchmarks, Google claims it outperforms its predecessor Gemini 2.5 Pro across multiple metrics while running three times faster. And these efficiency gains extend the token usage with the model using 30% fewer tokens on average than 2.5 Pro for everyday tasks while maintaining higher performance all the while.

Google is charging 50 cents per million input tokens and $3 per million output tokens. And for audio input, the cost remains at $1 per million tokens. And currently developers can access Gemini 3 flash through Google AI Studio, Vertex AI and Gemini CLI. So Gemini 3 flash is currently serving as the default mode in Gemini app, replacing 2.5 flash globally at no cost to users. Plus the model is also rolling out as the default for AI mode in Google search worldwide. And when it comes to capabilities, Google says that users can turn videos and images into actionable plans within seconds or build functional apps through voice dictation alone.

And this releases just Google’s wider expansion of the Gemini 3 family over the last month with the launch of Gemini 3 Pro and deep think mode reportedly processing over 1 trillion tokens daily through the API since then. But like and subscribe and check out this video here for more on the latest in AI news and thanks for watching. Bye. [tr:trw].