Portable solar generators are so useful that you increasingly see them at the beach, campground, job site, or anywhere without access to grid power. But they’re also an expensive luxury if shoved into storage as soon as you return home.
Technology
EcoFlow’s $200 PowerStream is so clever, you might buy a $4,000 solar generator
/cdn.vox-cdn.com/uploads/chorus_asset/file/25436922/IMG_4843.jpeg?ssl=1 "EcoFlow’s 0 PowerStream is so clever, you might buy a ,000 solar generator")
EcoFlow’s $200-ish PowerStream box can help maximize the year-round usefulness of the company’s own solar generators. It’s pitched as a plug-in “balcony solar system” that anyone can install, even if you’re renting an apartment.
These DIY systems from EcoFlow, Anker, and others became very popular around Europe over the last few years as energy costs soared amid steadily decreasing solar and battery prices. They can lower your energy bills and ensure a modicum of home backup in the event of a power outage.
Plug-in systems are built around a microinverter that feeds solar energy back into the home via a standard wall jack. The solar panels can be leaned up against a terrace wall, placed in a garden, or hung off a balcony railing. Any solar excess not used immediately by the home can be diverted into the solar generator’s big-ass battery for use later.
I’ve been testing an EcoFlow PowerStream setup in my own home, where I’ve installed $1,500 worth of EcoFlow’s portable solar panels on my roof, with the cables snaking through a window to the PowerStream box connected to a $2,599 EcoFlow Delta Pro. I also have six EcoFlow smart plugs attached to things like the washing machine, coffee maker, and home theater projector that tell the PowerStream to send more or less power.
It’s surprisingly simple to set up and get running, but not without some initial trepidation about safety, which I delve into below. I also made a mistake that killed the first PowerStream EcoFlow sent me. But importantly, I learned a lot about my family’s energy consumption habits and how much power 800W of haphazardly installed solar panels can produce under a weak northern sun.
So, it’s a shame that EcoFlow’s PowerStream is a Europe-only solution for now.
How we rate and review products
How does it save me money?
The PowerStream has three proprietary ports: one that connects to your MC4 solar panels; one that connects to your battery; and one AC output that plugs into a standard wall socket. The battery is entirely optional if you just want to feed every watt of solar power produced back into your home.
To understand how the PowerStream works, let’s look at a real example using screen recordings taken from the excellent EcoFlow app. Note how the direction of power delivery changes from GIF to GIF as EcoFlow’s smart plugs steadily increase demand.
In the first GIF above, I’m generating 397W of solar power, but my home is only demanding 290W — 250W for the “base load” and 40W for EcoFlow’s smart plugs (both of which I’ll describe in detail later). Since that load could be covered by solar power alone, the PowerStream didn’t need to request any power from the grid. It immediately sends the excess 107W of solar to charge the connected Delta Pro battery instead.
But after I turn on my home theater projector, the “smart plug” load jumps to 239W, and my solar panels can no longer cover the 489W now demanded by my home. So PowerStream diverts an extra 92W from the battery to cover the new total. Power grid demand is still 0W since the incoming solar power and battery power can cover the total demand of the home.
Finally, I turn on my Nespresso coffee machine, which activates the grid. The PowerStream has a maximum output of 800W (805W in this example), so it pulls an extra 845W from my grid provider to cover the 1,650W (1.4K plus 250W) my home is now demanding.
And while batteries do degrade if you cycle them every day like this, the LFP chemistry used by modern solar generators like the Delta Pro should maintain 80 percent of its charging capacity after 3,500 cycles — that’s almost 10 years. And it should still be good for 50 percent capacity after 6,000 cycles.
By the end of my example day, the PowerStream had produced a total of 3.03kWh, saving me a grand total of… 90 euro cents at my current energy rates. That might not sound like a lot, but at that rate, it will easily pay off the PowerStream after about a year of usage — faster in some climates and slower in others. Notably, it would have paid itself off in just months last year, when I was paying over three times as much per kWh due to global events.
Importantly, I have gained a degree of energy independence in this uncertain world — and making regular use of an expensive solar generator that was otherwise just waiting for the next road trip or natural disaster.
Great, but is feeding that much electricity into a standard wall socket actually safe?
A power outlet becomes a power inlet
It might seem odd and even unsafe to feed electricity into something called a power outlet, but AC wall jacks are, in fact, bidirectional under the right conditions.
The PowerStream is a mini power plant that automatically synchronizes with the grid to ensure seamless integration with the devices plugged into your home circuitry. Power flows from high to low voltages, which is why the grid voltage is slightly higher (by millivolts) than the devices plugged in. The PowerStream similarly alters its voltage to regulate the flow of power to your devices.
EcoFlow’s PowerStream is not a burden on 16A home circuits in Europe because it’s limited to a maximum of 800W (or 600W in Germany, currently) and requires less than 3.5A. Still, if the circuit is overloaded for any reason, the breaker will shut it off as usual, and the PowerStream will stop working.
And when there’s a power outage, the PowerStream will turn off automatically to ensure there’s no electricity in the wires in order to protect line workers from shock. The PowerStream will only turn back on when the grid power returns.
(This also means that the PowerStream, unlike a Tesla Powerwall or EcoFlow’s own Delta Pro Ultra home backup system, can’t automatically switch over to battery during a blackout to power the home. Instead you have to disconnect the battery — a 100-pound Delta Pro in my case — from the PowerSteam and wheel it to wherever power is needed, like the kitchen or basement.)
And despite how it looks in the app, the PowerStream isn’t actually sending power directly to the devices attached to those EcoFlow smart plugs. The PowerStream uses the smart plugs as signals to pump more or less juice into the stream of power entering the home, from which every device can drink.
EcoFlow says the PowerStream has obtained grid certification in over 10 countries, covering 70 percent of Europe. You might still have to register it with your local energy provider, however — so do check your local requirements. And the PowerStream isn’t available in the US at all due to restrictions that prohibit plug-in grid solutions.
Installation
The hardest part in setting up the PowerStream is doing the math to ensure your solar panels are wired up properly in series or parallel, especially when maxing out the system like I did. My bad math caused me to smoke the first PowerStream review unit by overvolting it (where, by “smoke,” I mean it just stopped working forever). Eventually, I figured out the right wiring configuration to keep everything under the 55V / 13A limit of the PowerStream’s solar inputs.
My setup is meant to be temporary because I still want to take the portable EcoFlow solar panels and Delta Pro on the road — but those panels need to be securely mounted. This can be done using EcoFlow’s own mounts (or your own, like I did) because a strong wind could easily send those lightweight panels flying. Of course, there’s nothing preventing you from connecting a PowerStream to rigid panels from any company you choose.
EcoFlow also sells its PowerStream with a new waterproof battery for a super tidy outdoor installation. Everything, including the PowerStream, is IP54 rated, and the 2kWh battery is even heated to ensure proper operation in temperatures down to -20 degrees Celsius (-4 degrees Fahrenheit). The kit runs completely silently, as does the PowerStream in my own comparatively crude installation.
With all the hardware installed, you then need to decide whether to prioritize power delivery from the PowerStream to your home circuitry or to battery storage. I currently default to home but would switch that to battery if a storm were in the forecast to ensure I had backup power. While you can schedule these modes to change throughout the day, you can only do it based on time, which is a shame. Ideally, it would automatically prioritize the battery if a big storm were in the forecast, as many home backup systems do.
Still, scheduling a change in priority modes can be very useful if you’re on a variable energy contract. That way, you can charge the battery during off-peak hours and then use that relatively cheap stored energy to power the home when electricity prices are highest.
1/8
The real trick to optimizing the PowerStream is to first determine your home’s idle baseline power usage — how many watts your home continuously pulls before turning on things like the dishwasher or coffee maker. Mine’s about 300W, according to the display on the smart energy meter in my utility closet. Ideally, EcoFlow’s PowerStream would get real-time readings from my smart meter, but that’s not currently possible.
So, I set what EcoFlow calls the base load to 250W in its app to ensure a 50W cushion. That way, I can avoid feeding energy back into the grid. Doing so has potential financial repercussions depending on your provider and where you live. As the sun returns here in the Netherlands, power companies are struggling to cope with oversupply under the current solar incentive scheme, resulting in charges levied against panel owners for energy returned to the grid.
EcoFlow smart plugs help direct traffic
Some PowerStream installations will benefit from a handful of Matter-enabled EcoFlow smart plugs installed on high-consumption devices like the TV, washer, dryer, dishwasher, refrigerator, and boiler to properly balance the load. They tell the PowerStream that those devices are demanding even more power than the base load, so it needs to ramp things up. Each smart plug costs €36 (about $39).
In the first screen recording below, you can see the PowerStream delivering 482W into the home. It determines this number by starting with the 250W I set as my base load and adding 232W to cover all the devices connected to the smart plugs. And since only 355W was coming in via solar at the time, it pulls an additional 127W from the connected Delta Pro battery for the home. In the second recording, the excess solar power is immediately diverted to the battery now that the projector is turned off.
And while the app shows the power grid delivering 0W into the home, this is pure fantasy since I don’t have smart plugs on every single device attached to my home. (And remember, my true baseline is closer to 300W.)
More power-heavy homes can just skip the smart plugs entirely. My maxed-out system with 800W of solar input can only generate a maximum of 4kWh per day, usually much less. That’s not enough to cover my daily base load of 6kWh (250W x 24 hours). So I should just feed enough solar power into the home to cover my base load during the day and send any excess to the EcoFlow battery for use when the sun goes down.
I have to say, I’m incredibly impressed by the PowerStream, especially now that it’s priced at just €150, almost half off the €279 it cost at launch last year. Not only does it help existing owners of EcoFlow’s solar generators maximize the value for their money but it also makes the idea of owning a solar generator more tempting knowing you can use it year-round, at home or off the grid.
It also makes the owner acutely aware of their energy habits. I’ve been obsessing over the EcoFlow app’s data like a new runner who just bought their first Garmin watch. For me, it’s been an interesting and relatively frugal first step toward energy independence. I now have real data and experience to help make a very complex decision about installing my own fixed panel system.
Photography by Thomas Ricker / The Verge
Continue Reading
Today on Decoder, I want to lay out an idea that’s been banging around my head for weeks now as we’ve been reporting on AI and having conversations here on this show. I’ve been calling it software brain, and it’s a particular way of seeing the world that fits everything into algorithms, databases and loops — software.
Software brain is powerful stuff. It’s a way of thinking that basically created our modern world. Marc Andreessen, the literal embodiment of software brain, called it in 2011 when he wrote the piece “Why software is eating the world” as an op-ed in The Wall Street Journal. But software thinking has been turbocharged by AI in a way that I think helps explain the enormous gap between how excited the tech industry is about the technology and how regular people are growing to dislike it more and more over time.
In fact, the polling on this is so strong, I think it’s fair to say that a lot of people hate AI. And Gen Z in particular seems to hate AI more and more as they encounter it. There’s that NBC News poll showing AI with worse favorability than ICE and only a little bit above the war in Iran and the Democrats generally. That’s with nearly two thirds of respondents saying they used ChatGPT or Copilot in the last month. Quinnipiac just found that over half of Americans think AI will do more harm than good, while more than 80 percent of people were either very concerned or somewhat concerned about the technology. Only 35 percent of people were excited about it.
Poll after poll shows that Gen Z uses AI the most and has the most negative feelings about it. A recent Gallup poll found that only 18 percent of Gen Z was hopeful about AI, down from an already-bad 27 percent last year. At the same time, anger is growing: 31 percent of those Gen Z respondents said they feel angry about AI, up from 22 percent last year.
Now, I obviously talk to a lot of tech executives and policy people here on Decoder, and I will tell you, they all know AI isn’t popular, and they can all see how that’s playing out in real life. Here’s Microsoft CEO Satya Nadella talking about how the tech industry needs to make the case for the investments it’s making in AI:
Satya Nadella: At the end of the day, I think this industry, to which I belong, needs to earn the social permission to consume energy because we’re doing good in the world.
I think it’s safe to say that the tech industry and AI have not earned any of that social permission yet. Politicians from both sides of the aisle are opposing data center buildouts. Politicians in local communities that support data centers are getting voted out of office. And in the most depressing reminder of how much political violence has become a part of everyday American life, politicians who’ve supported data centers have had their houses shot at. OpenAI CEO Sam Altman has had Molotov cocktails thrown at his house.
It’s sad that I’m going to have to say this again on the show, and it’s sad that we’re going to have commenters who disagree, but this violence is unacceptable. If you want to meaningfully oppose AI in a way that lasts, you should speak loudly with your dollars in the market and your attention online, and you should speak loudly with your votes. You should participate in a democratic regulatory and political process. Anything else will get dismissed and perpetuate the cycle. That dismissal is already happening.
I also think it’s incredibly important for our politicians and tech executives to make sure our political process makes people feel empowered, not helpless, which is a specific kind of nihilism they have all greatly contributed to. The violence is a result of that helplessness and nihilism. And the most powerful people in our society ought to reckon with that, especially as they run around saying AI will wipe out all the jobs. I’m not even exaggerating this. Here’s Anthropic CEO Dario Amodei saying he thinks AI will wipe out all the jobs:
Dario Amodei: Entry-level jobs in areas like finance, consulting, tech and many other areas like that —- entry-level white-collar work — I worry that those things are going to be first augmented, but before long replaced by AI systems. We may indeed —- it’s hard to predict the future — but we may indeed have a serious employment crisis on our hands as the pipeline for this early-stage, white-collar work starts to contract and dry up.
What I see when I encounter clips like this is the true gap between the tech industry and regular people when it comes to AI — and also the limit of software brain. Like I said, everyone in tech understands how much regular people dislike AI. What I think they’re missing is why. They think this is a marketing problem. OpenAI just spent $200 million on the TBPN podcast because the company thinks it will help make people like AI more. Sam Altman has said so explicitly:
Sam Altman: Oh, they are genius marketers and I would love to have better marketing. Somebody said to me recently that if AI were a political candidate, it would be the least popular political candidate in history. And given the amazing things AI can do, I think there’s got to be better marketing for AI.
It feels like someone just needs to say this clearly, so I’m just going to do it. AI doesn’t have a marketing problem. People experience these tools every single day. ChatGPT has 900 million weekly users, trending to a billion, and everyone has seen AI Overviews in Google Search and massive amounts of slop on their feeds. You can’t advertise people out of reacting to their own experiences. This is a fundamental disconnect between how tech people with software brains see the world and how regular people are living their lives.
Image: The Verge
So what is software brain? The simplest definition I’ve come up with is that it’s when you see the whole world as a series of databases that can be controlled with structured language and software code. Like I said, this is a powerful way of seeing things. So much of our lives run through databases, and a bunch of important companies have been built around maintaining those databases and providing access to them.
Zillow is a database of houses. Uber is a database of cars and riders. YouTube is a database of videos. The Verge’s website is a database of stories. You can go on and on and on. Once you start seeing the world as a bunch of databases, it’s a small jump to feeling like you can control everything if you can just control the data.
But that doesn’t always work. Here’s an example: Elon Musk and DOGE showed up in the government, and the first thing they did was take control of a bunch of databases. And they ran into the undeniable fact that the databases aren’t reality, and DOGE ended in hilarious failure. It turns out software brain has a limit, and the government isn’t software. People aren’t computers, and they don’t live in automatable loops that can be neatly captured in databases.
Anyone who’s actually ever run a database knows this. At some point, the database stops matching reality. And at that point, we usually end up tweaking the database, not the world. The AI industry has fully lost sight of this. AI thrives on data. It’s just software. And so the ask is for more and more of us to conform our lives to the database, not the other way around.
Let me offer you another example that I think about all the time, especially as AI finds real fit as a business tool. It’s the idea that AI is coming for lawyers and the legal system. The AI industry loves to talk about not needing lawyers anymore, which is already getting all kinds of people into all kinds of trouble. But I get it. I’ve spent a lot of time with lawyers. I used to be a lawyer. My wife is still a lawyer. Some of my best friends are lawyers.
Verge subscribers, don’t forget you get exclusive access to ad-free Decoder wherever you get your podcasts. Head here. Not a subscriber? You can sign up here.
I also spend all of my time at work talking to tech people. And so over time, I’ve learned that the overlap between software brain and lawyer brain is very, very deep. Alluringly deep. If the heart of software brain is the idea that thinking in the structured language of code can make things happen in the real world, well, the heart of lawyer brain is that thinking in the structured legal language of statutes and citations can also make things happen. Hell, it can give you power over society.
There are other commonalities. Both software development and the law depend heavily on precedent. We have a body of case law in this country, and we use it over and over again to help us resolve disputes. Much like software engineers have libraries of code that they turn to repeatedly to build the foundations of their products. I can go on.
At the end of the day, both lawyers and engineers do their best to use formal, structured language to guide the behavior of complicated systems in predictable and potentially profitable ways. I am far from the first person with this idea. Larry Lessig wrote a book called Code and Other Laws of Cyberspace in 2000. It’s just as relevant today as it was a quarter century ago.
And so you have this intoxicating similarity between law and code, and it trips people up all the time. People are constantly trying to issue commands to society at large like it’s a computer that will obey instructions. There are examples of this big and small. My favorite are those Facebook forwards insisting Mark Zuckerberg does not have the right to publish people’s photos. Honestly, I look at these, and I think it would be great if the law was actually code. Maybe things would be more predictable. Maybe we’d feel more in control.
But law isn’t actually code, and society and courts aren’t computers. I have to remind our fairly technical audience on Decoder and at The Verge all the time that the law is not deterministic. You simply cannot take the facts of a case, the law as written, and predict the outcome of that case with any real certainty, even though the formality of the legal system makes people think it works like a computer, that it’s predictable.
Because at the end of the day, it’s actually ambiguity that’s at the very heart of our legal system. It’s ambiguity that makes lawyers lawyers. Honestly, it’s ambiguity that makes people hate lawyers because it’s always possible to argue the other side, and it’s always possible to find the gray area in the law. That’s why prosecutors end up working as defense attorneys and why our regulators tend to end up working for big corporations.
So you can see the obvious collision between software brain and lawyer brain. This thing that looks like a computer isn’t actually anything at all like a computer. A lot of people even argue that the law should be more like a computer, that the system should be verifiable and consistent, and that merely issuing the right commands at the right times should lead to objectively correct outcomes.
Bridget McCormack, who used to be the chief justice of the Michigan Supreme Court, was on Decoder a few months ago pitching a fully automated AI arbitration system. Her argument to me was that people perceive the traditional legal system to be so unfair, they will accept a worse outcome from an automated system as more fair as long as they feel heard. And if there’s one thing AI can do, it’s sit there and listen all day and night. I don’t know if any of that is correct or even workable, but I do know software brain, and that is pure software brain. The idea that we can force the real world to act like a computer and then have AI issue that computer instructions.
You can see the same thing happening in every other kind of industry. You don’t hire a big consulting firm to actually come in and study your business and make it more efficient. You hire them to make slide decks that justify layoffs to your board and shareholders. Big consulting firms are great at this, and now they’re just going to generate those decks with AI. They are already doing this and the layoffs have already begun.
Any business process that looks like code talking to a database in a repetitive way is up for grabs. That’s why Anthropic has been so relentlessly focused on enterprise customers, and it’s why OpenAI is now pivoting to business use. There’s real value in introducing AI to business because so much of modern business is already software, collecting data, analyzing it, and taking action on it over and over again in a loop. Businesses also control their data, and they can demand that all their databases work together. In this way, software brain has ruled the business world for a long time. And AI has made it easier than ever for more people to make more software than ever before, for every kind of business to automate big chunks of itself with software. The absolute cutting edge of advertising and marketing is automation with AI. It’s not being in creative.
But not everything is a business, not everything is a loop, and the entire human experience cannot be captured in a database. That’s the limit of software brain. That’s why people hate AI. It flattens them. Regular people don’t see the opportunity to write code as an opportunity at all. The people do not yearn for automation. I’m a full-on smart home sicko; the lights and shades and climate controls of this house are automated in dozens of ways. But huge companies like Apple, Google and Amazon have struggled for over a decade now to make regular people care about smart home automation at all. And they just don’t.
AI isn’t going to fix that. Most people are not collecting data about every single thing that they do. And if they’re collecting any at all, it’s stored across lots of different systems — your email in Gmail, your messages in iMessage, your work schedule in Outlook, your workouts in Peloton. Those systems don’t talk to each other and maybe they never will, because there’s no reason for them to. And asking people to connect them all freaks them out.
Even taking the time to consider how much of your life is captured in databases makes people unhappy. No one wants to be surveilled constantly, and especially not in a way that makes tech companies even more powerful. But getting everything in a database so software can see it is a preoccupation of the AI industry. It’s why all the meeting systems have AI note takers in them now. It’s why Canva, which is design software, now connects to corporate email systems. My friend Ezra Klein just went to Silicon Valley, and he described the people that are actively trying to flatten themselves into a database:
Ezra Klein: You might think that A.I. types in Silicon Valley, flush with cash, are on top of the world right now. I found them notably insecure. They think the A.I. age has arrived and its winners and losers will be determined, in part, by speed of adoption. The argument is simple enough: The advantages of working atop an army of A.I. assistants and coders will compound over time, and to begin that process now is to launch yourself far ahead of your competition later. And so they are racing one another to fully integrate A.I. into their lives and into their companies. But that doesn’t just mean using A.I. It means making themselves legible to the A.I.
You can give it access to everything that’s there: your files, your email, your calendar, your messages. It operates continuously in the background, building a persistent memory of your preferences and patterns so it can better act on your behalf. The cybersecurity risks are glaring, but there’s a reason millions of people are using it: The more of your life you open to A.I., the more valuable the A.I. becomes.
I’ve reviewed a lot of tech products over the past decade and a half, and all I can tell you is that it is a failure when you ask people to adapt to computers. Computers should adapt to people. And asking people to make themselves more legible to software, to turn themselves into a database, is a doomed idea. It’s an ask so big, I can’t imagine a reward that would make it worth it for anyone, even if the tech industry wasn’t constantly talking about how AI will eliminate all the jobs, require a wholesale rethinking of the social contract and — oops — also the latest models might cause catastrophic cybersecurity problems that might lead to the end of the world.
Does this sound like a good deal to you? Can you market your way out of this? This only makes sense if you have software brain, if your operative framework is to flatten everything into databases that you can control with structured language. The people paying thousands of dollars a month to set up swarms of OpenClaw agents and write thousands of lines of code, they’re people who look at the world and see opportunities for automation, to repeat tasks, to collect data, to build software. AI is great for them. It’s even exciting in ways that I think are important and will probably change our relationship to computers forever.
For everyone else, AI is just a demanding slop monster. It’s a threat. I’m not saying regular people don’t use Excel or Airtable to plan their weddings or have fun throwing PowerPoint parties, or even that AI won’t be useful to regular people over time. I think a lot of people enjoy data and tracking different parts of their lives. There’s my WHOOP band. I’m just saying these things aren’t everything. Not everything about our lives can be measured and automated and optimized. It shouldn’t be.
And so the tech industry is rushing forward to put AI everywhere at enormous cost — energy, emissions, manufacturing capacity, the ability to buy RAM — and locked into the narrow framework of software brain without realizing they are also asking people to be fundamentally less human. They then sit around wondering why everyone hates them. I don’t think a couple haircuts are going to fix it.
Questions or comments about this episode? Hit us up at decoder@theverge.com. We really do read every email!
Decoder with Nilay Patel
A podcast from The Verge about big ideas and other problems.
SUBSCRIBE NOW!
Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.
NEWYou can now listen to Fox News articles!
Most people think of Toyota and picture a Camry, a Tacoma, maybe a Prius. A 7-foot-2 robot shooting free throws at halftime of a professional basketball game? That’s a harder image to conjure. But recently, that’s exactly what happened at Toyota Arena Tokyo, and around 8,400 fans watched it go down live.
The robot is called the CUE7. It smoothly stood up from a seated position, dribbled a basketball and sank a free throw without any human input. The crowd applauded. The engineers probably exhaled. Toyota had officially debuted its most advanced AI-powered humanoid robot, and it chose basketball as the venue.
So why is a car company building basketball robots? And what does any of this have to do with you? More than you might think.
Sign up for my FREE CyberGuy Report
- Get my best tech tips, urgent security alerts and exclusive deals delivered straight to your inbox.
- For simple, real-world ways to spot scams early and stay protected, visit CyberGuy.com – trusted by millions who watch CyberGuy on TV daily.
- Plus, you’ll get instant access to my Ultimate Scam Survival Guide free when you join.
AI-POWERED ROBOT SINKS SEEMINGLY IMPOSSIBLE BASKETBALL HOOPS
Toyota’s CUE7 robot handles the ball with precision, showing how AI can learn complex physical movement. (Toyota Motor Corporation)
The CUE7 started from scratch, on purpose
Here’s the thing that makes the CUE7 genuinely different from its predecessors: Toyota’s team discarded everything they had built and started over.
“We made full use of AI, and we discarded everything we had built up and started again from scratch,” said Tomohiro Nomi, research leader for humanoid robots at Toyota’s Frontier Research Center.
That’s not a small statement. The CUE series goes back to 2017, when a group of Toyota employees launched it as a voluntary side project on their own time. It eventually became an official research program, and over nearly a decade, the team stacked up some genuinely impressive hardware. The CUE3 earned a Guinness World Record in 2019 for most consecutive basketball free throws by a humanoid robot (assisted), sinking 2,020 in a row. Then the CUE6 earned the record for the farthest basketball shot by a robot, connecting from about 80 feet 6 inches) away.
So the legacy was already there. What changed with CUE7 was the philosophy behind how it learns.
From human programming to AI that figures it out alone
Earlier versions of the CUE relied on something called model predictive control. Basically, human engineers programmed exactly how the robot should move, step by step. It worked well enough to break world records. But it also had a ceiling. Every new motion required new programming by a human being.
The CUE7 instead uses reinforcement learning powered by artificial intelligence. It learns to shoot the ball based on its own experience and trial and error rather than pre-programmed instructions. The AI acts as an autonomous agent: it tries something, observes the result, adjusts and tries again. Over enough repetitions, it gets good. Really good.
The hybrid control system merges reinforcement learning with model predictive control, creating a robot that adapts to unexpected situations rather than just following a fixed script. Think of it as the difference between a player who memorized every play in the book and one who reads the game in real time. CUE7 is learning to read the game.
What’s actually inside the CUE7 robot
The CUE7 stands about 7 feet 2 inches tall and weighs roughly 163 pounds, making it about 40% lighter than the previous version, which came in around 265 pounds. Toyota pulled that off by simplifying the structure and reducing the number of axles.
It also switched from four wheels to two, which makes its movement faster and more fluid. One moment that really stood out was how smoothly it can rise from a seated position. That kind of motion, especially at this size, takes serious engineering and drew a reaction from a crowd of more than 8,000 people.
For sensing and aiming, the robot uses lidar sensors in its torso to detect its surroundings, along with a stereo camera in its head to calculate distance and angle. It is powered by high-performance batteries adapted from Toyota’s racing tech.
Here’s where it gets interesting. The robot measures the distance to the hoop, calculates the angle, determines the right trajectory and then releases the shot with controlled force. If it misses, it learns from that attempt and adjusts on the next one.
ROBOT PLAYS TENNIS WITH HUMANS IN REAL TIME
During a live game demo, the robot lines up a shot, highlighting how machines can adapt in real-world environments. (Toyota Motor Corporation)
The AI that actually makes this work
Toyota trained the system using human motion data, which is what gives CUE7 its surprisingly natural movement. Rather than looking mechanical, its actions mirror how a person actually moves, and that’s by design.
That same combination of real-time calculation and learned experience is what lets it handle something like dribbling (fluid, continuous) alongside shooting (precise, calculated) without the two working against each other.
Toyota says testing that kind of learning in a live environment is a key part of the project.
“We believe it is an exceptionally valuable opportunity to validate a reinforcement-learning-based robot in the inherently uncertain environment of a basketball arena,” Tomohiro Nomi, Head of Humanoid Robotics Research Unit, Frontier Research Center, Toyota Motor Corporation, told CyberGuy. “Moving forward, we will continue developing robots that inspire and bring joy to people.”
What this means to you
You’re probably not buying a robot basketball player anytime soon. But here’s the part worth paying attention to: the same AI that helps CUE7 sink free throws is the technology Toyota is actively developing for manufacturing, automotive systems and real-world robotics.
Basketball demands everything that manufacturing robots struggle with: target identification, distance gauging, trajectory computation, coordinated movement and precise force control, all in sequence and under pressure. Toyota chose basketball specifically because it tests all those capabilities at once, in an environment where success and failure are completely obvious.
The reinforcement learning powering CUE7 could eventually show up in factory robots that adapt mid-shift when production requirements change, in vehicles that handle unexpected road conditions more fluidly, or in home and care robots that need to navigate unpredictable environments. Toyota treats CUE7 as a testbed for vision systems, motion control and coordinated movement, with capabilities that reach well beyond halftime demonstrations into broader real-world applications.
When Toyota teaches a robot to play basketball, it’s really teaching machines how to learn. And that skill transfers. In other words, this is less about basketball and more about teaching machines how to learn physical skills in unpredictable environments. That is where the real impact starts to show up.
THE NEW ROBOT THAT COULD MAKE CHORES A THING OF THE PAST
CUE7 sinks a free throw, a simple moment that reflects a bigger shift toward AI that learns through experience. (Toyota Motor Corporation)
Take my quiz: How safe is your online security?
Think your devices and data are truly protected? Take this quick quiz to see where your digital habits stand. From passwords to Wi-Fi settings, you’ll get a personalized breakdown of what you’re doing right and what needs improvement. Take my Quiz here: Cyberguy.com.
Kurt’s key takeaways
The CUE7 is a fascinating piece of technology, but the real story isn’t about basketball. It’s about a fundamental shift in how robots are trained, moving away from rigid human programming toward AI systems that learn through experience and adapt on the fly. What started as a voluntary employee side project in 2017 has grown into a genuine proving ground for Toyota’s embodied AI research. Nearly a decade in, the results are landing in front of thousands of live spectators and stacking up Guinness World Records along the way. The CUE7 made a free throw at halftime in front of a packed arena. More importantly, it demonstrated that AI-powered machines can now acquire complex physical skills through trial and error, the same basic way humans do. That’s a shift with implications that reach far beyond the basketball court.
If a robot can teach itself to make free throws better than most humans ever will, purely through AI-driven trial and error, what physical skill do you still believe machines will never be able to learn on their own? Let us know by writing to us at Cyberguy.com.
CLICK HERE TO DOWNLOAD THE FOX NEWS APP
Sign up for my FREE CyberGuy Report
- Get my best tech tips, urgent security alerts and exclusive deals delivered straight to your inbox.
- For simple, real-world ways to spot scams early and stay protected, visit CyberGuy.com – trusted by millions who watch CyberGuy on TV daily.
- Plus, you’ll get instant access to my Ultimate Scam Survival Guide free when you join.
Copyright 2026 CyberGuy.com. All rights reserved.
Technology
The Iranian women Trump ‘saved’ from execution are simultaneously real and AI-manipulated
Only the night before, he had posted on Truth Social about the imminent executions of these women, quoting a screenshot that included a collage of eight glamorously backlit, soft-focus portraits. The photos of the women were immediately accused of being AI-generated. “Trump is begging Iranian leaders to not execute 8 AI-generated women. This is the funniest thing I’ve ever seen,” said one viral X post.
On top of that, almost immediately after Trump’s announcement, Mizan, an Iranian state news agency, called the president a liar. “Last night, Donald Trump, citing a completely false news story, called on Iran to overturn the death sentences of eight women.” Mizan said that some of the women had already been released and others were facing prison time but not execution, and furthermore said that Tehran had made no concessions — presumably, the status of the women has not changed.
The X account for the Iranian embassy in South Africa, perhaps the most relentless shitposter among Iran’s state-affiliated accounts, was quick to pile on by generating its own set of eight women:
The collage that Trump posted is, at the very least, AI-modified, Mahsa Alimardani, the associate director of the Technology Threats & Opportunities program at WITNESS, told The Verge. But the women themselves are real. The woman in the top right corner of the collage is Bita Hemmati, whose photograph appeared in several news stories in various right-leaning news outlets last week. Hemmati is confirmed to have received a death sentence issued by Branch 26 of the Tehran Revolutionary Court for “operational action for the hostile government of the United States and hostile groups.”
Alimardani named six of the women (Bita Hemmati, Mahboubeh Shabani, Venus Hossein-Nejad, Golnaz Naraghi, Diana Taherabadi, Ghazal Ghalandri), and said that the identities of the final two (said to be Panah Movahedi and Ensieh Nejati) were still unverified. The six verified women participated in protests against the government in January. Aside from Hemmati, none of the other women are reported to have received death sentences.
It’s not surprising that Trump has a careless disregard for the truth; it’s not surprising, either, for the Iranian regime to fudge the details to suit its own narrative, or to make light of real political prisoners in order to dunk on the United States.
The additional wrinkle is that the account mocking Trump for coming to the rescue of “8 AI-generated women” is the very same one that landed South Korean president Lee Jae-myung in hot water when he quoted a misleading labeled video posted by that account. Israeli officials have accused the account of being “well-known for spreading disinformation.” The case of the sketchy Lee Jae-myung quote-post is a story of mingled truth and misinformation, where the post got facts very wrong, but the video — of Israeli Defense Forces soldiers shoving a limp body off a rooftop in Gaza — was real, documenting an event that possibly implicates Israeli forces in a violation of international law.
The case of the eight Iranian protesters also features that same mingling of fact and fiction into a fuzzy distortion that fuels an endless disputation of real human rights violations. Their lives have been reduced to glossy pixels and quote-dunks, the stuff of propaganda and parody. While known liars fight with each other on the internet about who these women are and what will happen to them, they — verifiably six of them, at least — remain real people who exist beyond the Iranian internet blackout.
-
Vermont2 minutes agoLetter to the Editor: A different path for Vermont’s environmental future
-
Virginia8 minutes agoWhy the Virginia redistricting referendum wasn’t a slam dunk for Democrats
-
Washington14 minutes agoSpringtime in Washington means it’s time for another round of federal privacy legislation | Brookings
-
Wisconsin20 minutes agoWhat can and can’t you recycle in Wisconsin? Here are the rules to know
-
West Virginia26 minutes agoThis Is The Friendliest Small Town in West Virginia
-
Wyoming32 minutes agoGovernor Mark Gordon Discusses Water-saving Measures By Data Centers In Southeast Wyoming
-
Crypto38 minutes agoBen McKenzie is Still Mad at Matt Damon For Those Crypto Ads
-
Finance44 minutes agoAuto Finance Capital Summit | Insights | Mayer Brown