Chinese automakers are following Tesla’s bet that robots are the next big profit machine
4:24 PM PDT · August 28, 2026
The hype around humanoid robots isn’t particularly new. Thank Tesla CEO Elon Musk and his Optimus robot, as well as the myriad videos of Boston Dynamics’ Atlas robot, for that.
Behind that hype, though, there is real progress. The physical capabilities of robots continue to improve, and researchers now believe that the AI techniques behind large language models can make complex robots capable of learning nearly any task.
Those tailwinds have encouraged a new batch of companies to jump in on the promise of profits from humanoid robots. And many of the latest entrants are Chinese automakers.
Earlier this week, Xpeng’s robotics unit raised more than $900 million at a post-money valuation of more than $6.3 billion. The round, led by IDG Capital with participation from Gaorong Ventures, Tencent, and Alibaba, was described by the company as the largest single-round private financing ever recorded in China’s “embodied AI” industry (AI systems built directly into physical machines).
This month, AiMOGA, the robotics unit of China’s Chery Automobile, reportedly began preparing for an IPO , while BYD unveiled a humanoid robot called Xiao Di. Other Chinese automakers, including Changan, GAC, Li Auto, SAIC, and Seres, are also developing humanoid robots.
Among all of them, Xpeng is the Chinese automaker that most closely watches and follows Tesla’s initiatives, according to Michael Dunne, CEO of San Diego- and Singapore-based advisory firm Dunne Insights.
“It’s the most focused on autonomy, it’s the first to commit in a big way to humanoid robots,” Dunne told TechCrunch, adding that Xpeng founder He Xiaopeng is a tech billionaire known for his agility and quick adjustments. “He sees razor-thin profit in cars on the near horizon. Robots look much more promising.”
Xiaopeng and Xpeng co-president Brian Gu are bullish enough that they’ve put their own funds behind the robotics unit. According to the WSJ, the pair invested about $100 million into the recent fundraising round.
Xpeng’s bet is on Iron, a humanoid robot with a realistic human shape that is built for commercial deployment.
Chinese automakers like Xpeng do bring a manufacturing edge.
“They have all the hardware to get the job done,” Dunne said. “Question is if they can catch Tesla on the AI side if the equation.”
There are, of course, many other companies developing humanoid robots, including Agility Robotics, Apptronik, and Figure, all chasing the same goal: commercial deployment at scale.
Hyundai-owned Boston Dynamics is getting closer to that goal. Hyundai plans to bring Boston Dynamics’ Atlas humanoid robot to its Georgia factory this year and eventually deploy the robots for tasks like parts sequencing by 2028. The Korean automaker, which partnered with Google’s AI research lab DeepMind to speed up the development of Atlas, is opening a U.S. facility this year called a Robot Metaplant Application Center, which will teach robots how to map movements like lifts and turns.
Other automotive companies are also jumping, including supplier Mobileye, which acquired humanoid robot startup Mentee Robotics earlier this year for $900 million. Even Rivian is dabbling in robots with its Mind Robotics spinout — although its robots are not expected to look quite like the humanoids in development elsewhere.
Is the best way to watch a movie on a pair of sunglasses?
3:42 PM PDT · August 28, 2026
I am nothing if not a huge movie buff. I watch way too many of them, and I’m always on the hunt for a new format in which to experience them. So when XREAL, the smart glasses company, sent me an a01 — one of its newer models, which it released in May of this year — I was eager to give them a spin as the newest vector by which to satisfy my media fanaticism.
The a01 isn’t a particularly sophisticated smart glasses model. Unlike more software-heavy AR glasses like, say, the Meta Orion or Snap’s Specs, it’s basically just an external monitor. It also doesn’t have a battery or an internal power mechanism. Instead, a simple USB-C cable plugs the glasses into a device of your choosing, which then becomes the headset’s power source. It’s also not so expensive, at an accessible price point of around $300.
The a01 is actually optimized for gaming, in that it can be plugged into a Steam Deck or other handheld gaming device. However, XREAL also advertises them as a way to watch movies and TV — and since that’s more my speed, once I had the a01 in hand, I plugged it into my personal laptop and booted up the Criterion Channel. I then sat watching David Lynch’s film “Wild at Heart” for a while, enjoying a sequence where Nicolas Cage, dressed in a snakeskin jacket, beats up a guy in a bar and then sings an Elvis song.
I’ll say this: The images look quite good. The glasses, which come outfitted with dual mini OLED panels, provide quite a nice image (those panels offer a 1080p resolution with up to 1,600 nits of brightness), with very vibrant colors. If you aim the glasses at a wall, it feels vaguely like you’re using a really vivid home projector — or perhaps are at a drive-in movie. In a dark room, you’re one step closer to the in-theater experience.
However, the overall experience also brought some questions to mind. Namely, why would I sit next to my computer with glasses on my head watching a thing that is also playing on my laptop only 14 inches away? The reason, XREAL offers, is that the glasses are more immersive (they claim the device’s projections are equivalent to viewing content on a 147-inch screen). Still, the redundancy of watching a movie while it plays right next to you makes you question what the actual purpose of the device is.
XREAL has suggested that the glasses can function as a “second monitor” (indeed, they’ve actually been referred to as a “wearable display”) but the functionality of this is, again, questionable. It’s rather difficult to see anything other than what the glasses are projecting — which would make it quite difficult to, say, work on a laptop while also wearing them.
The glasses can also be connected to your phone. Unfortunately, I have an older iPhone, which means that the a01’s cable is not compatible with the phone’s port. An adapter would have been necessary to link the two.
The user experience is easy enough to imagine, however. Connecting the glasses to a phone means that you’re not captive to an indoor experience anymore as you would be with a heftier device like a laptop. You can watch a movie while you’re traveling on a plane or a train (or, hell, while you’re walking down the street — although I would say this last option is generally ill-advised unless you want to accidentally walk into traffic).
However, there are still inconvenient limitations with using the device this way. For one thing, the glasses still only function as a screen-mirroring device — meaning that the screen of your phone needs to remain active while you’re using the glasses. This brings us back to the redundancy problem. You’re watching a video on a screen attached to your face while the same video plays on a different screen that is located less than a foot away. You could partially solve this issue by putting the phone in your pocket, but the chances seem high that any jostling might upset the device’s playback functionality.
It’s worth noting that the device can also be paired with a separate device, dubbed the Beam Pro, which is essentially a mini-tablet and can act as an isolated streaming hub. Users download shows and movies onto the Beam, connect it to the glasses, and watch. However, this device will cost you another $200.
Then there’s the heat. It doesn’t take long for the a01 to start warming up — producing an odd tingling sensation on the bridge of your nose and over your eyes. This is, of course, not an experience unique to XREAL’s products — it’s a well-known defect of most XR glasses. You can only cram so much computing into a small plastic device before all the electrical processing begins to warm everything up. Still, it’s a tad disconcerting, and not exactly what you would want from an accessory that you’re wearing on your face.
I will say that — heat aside — the a01 is a relatively lightweight and comfortable device — and it isn’t overly cumbersome like other smart glasses that I’ve worn. (Having given Snap’s Specs a try at CES earlier this year, I promise you those are significantly heavier — although it’s also a very different kind of device than the a01.)
XREAL continues to iterate its product line, with each new device seeming to improve upon the previous one. Indeed, some of the existential dilemmas present in the a01 and previous XREAL headsets seem to have been ironed out in the company’s newest (and yet to be released) device: Project Aura — which I caught a glimpse of during my visit to Google I/O earlier this year — promises a significantly more immersive and convenient experience.
The Aura is powered by Android XR, an extended reality operating system developed by Google and Samsung. The Aura comes with native hand tracking (which is absent in the a01), as well as access to the Google Play Store, giving the glasses significantly more interactive abilities and AR potential. It also comes with a puck, tethered to the glasses, that acts as both the charging source and a compute node. The puck, which can be easily placed in your pocket, means that — unlike the a01 — you have substantially more mobility and you don’t have to keep it plugged into a separate device.
Let’s return to the a01, though. Unfortunately, from a cinephile’s perspective, watching movies on a pair of sunglasses just isn’t ideal. In general, movie fans like a big screen — the bigger, the better, really. In a world of 4K OLEDs of varyingly gargantuan sizes, consumers have a lot of options. My TV — a 55-inch TCL S-series — isn’t even a particularly powerful device, but it provides a home-viewing experience that is more comfortable and satisfying than what the a01 can provide. To my mind, watching a movie at home on a large flat screen is second only to actually going to a theater. Watching a film on tiny screens less than an inch from your eyes, meanwhile, is an interesting experience for its distinct sense of immersion but not what I’d call optimal.
The a01 is an interesting glimpse into a hardware industry that continues to evolve and that is still finding its footing with consumers. I’m curious to see how the user experience shifts with XREAL’s upcoming Aura, and I’m game to reevaluate my movie-watching preferences when that time comes.
An Anthropic researcher just gave us a peek at self-improving AI
12:30 PM PDT · August 28, 2026
Training AI models with other AI models has become a very popular goal for neolabs — and now, a researcher in Anthropic’s fellows program has given us an early look at what it might look like in practice.
On Friday, Anthropic published a new paper titled “ Automated Researchers Can Reliably Mitigate Alignment Failures,” detailing how AI systems could reliably improve a model’s performance on a set of alignment benchmarks. When given 10 benchmarks for specific misaligned behaviors, the automated systems were able to improve performance on every single one without degrading overall performance.
Led by Anthropic fellow Chen Yueh-Han, the system replicates much of the traditional approach to research. Each automated system searches the available literature, proposes a method, and trains the model using that method for 30 minutes, gradually increasing the benchmark over several iterations. Effective methods are preserved while ineffective ones are discarded, allowing the system to operate quickly and at a great scale.
“Overall, these results provide early evidence that automated alignment post-training could become practical in the near term,” the paper reads.
The paper is a step toward recursive self-improvement, which many see as the next significant step in AI progress. If models can improve their own alignment training, it’s plausible they could improve training practices more broadly — at which point, human AI researchers might soon become obsolete.
The paper isn’t shy about addressing this idea, explicitly comparing the Automated Alignment Researcher (AAR) to its human equivalent. “The best AAR method beats what experienced humans propose, on average within six hours,” the paper reads. “Human guided research directions do not lead to stronger performance.”
There’s even a cost comparison, in case anyone wasn’t convinced. “An AAR costs roughly $4 per hour in API inference against the $150 per hour we pay our human researchers.”
In fairness, the paper also points out a few limitations to this approach. The automated system only works insofar as the benchmarks reflect the actual alignment goals, and even then there’s significant work to be done in establishing and maintaining those benchmarks — not to mention maintaining and expanding on the literature the automated researchers are drawn from.
Brave’s browser one-ups Chrome with its new support for email aliases
11:50 AM PDT · August 28, 2026
Chrome alternative Brave just added a good reason to make the switch to its browser: email aliases. The feature, announced this week, allows Brave’s users to sign up for websites and other online services without having to share their personal email addresses.
To use the feature, you first need to create an account with Brave, where you can provide your real email address. Then, once logged into the browser, you’ll see a pop-up appear when clicking an email field on a website, offering the option to use an alias instead. (If the pop-up doesn’t appear, just right-click on the field to access the option.) The email alias will be filled in on the website’s form, and any emails coming to that address will be forwarded to your primary email account.
The company explains that using an alias can protect your privacy, particularly because websites use your email as a personal identifier, allowing them to target you for ads. Ad tech giants like Meta are also able to match the email you provided to a retailer’s website, for instance, with the address they have on file for you, allowing these companies to track what you bought or viewed.
Plus, if any website where you shared your email gets hacked, the hackers will then know your primary email address, and this information can be leaked and shared with data brokers.
Brave says it built the new feature to plug this privacy hole, adding that it doesn’t read the contents of emails sent to an alias — it only processes them for spam and virus filtering. After the email is forwarded, Brave deletes it from its servers. All data in users’ Brave accounts is also encrypted at rest, and any notes you save alongside an alias are stored locally on your device, unless you turn on the sync option. If syncing is enabled, notes will be encrypted end-to-end.
As the email alias feature is still new, Brave cautioned that some of the forwarded mail may initially go to your spam folder, but this should improve as Brave builds up its reputation score as a mail provider.
For the time being, Brave users are being offered five free email aliases. More will be available for Premium plan users in the future.
Open-weight AI companies are the Valley’s hottest acquisition targets
11:19 AM PDT · August 28, 2026
Everyone’s waiting for Nvidia to confirm this week’s most interesting tech deal: A reported $13 billion acquisition of Hugging Face, a platform for sharing open-weight AI models and benchmarks.
Now best known as the target for a team of reward-hacking OpenAI agents, Hugging Face is at the center of the ecosystem of developers building and deploying LLMs that aren’t owned by frontier labs. Think of it as a kind of GitHub for the AI era.
Rumors of that deal come after Nvidia struck a $6 billion agreement with Poolside, an open-weight model builder, that will see most of its employees move to the chip-making giant. And two weeks ago, Stripe acquired OpenRouter, the top provider of open-weight models to businesses, for more than $7 billion.
That’s a lot of capital pouring into a sector based on giving stuff away, and it reflects the latest trends in the AI sector.
For Nvidia, there’s a need to avoid further dependence on its deals with the major hyperscalers and frontier labs. That’s particularly the case when major AI model builders like OpenAI and Google are also building their own inference chips, like OpenAI’s Jalapeño, whose capabilities were announced this week. If model builders are making chips, Nvidia wants a chunk of the model-making business.
Nvidia already builds its own Nemotron family of open-weight models, but their uptake hasn’t been huge. By taking control of the largest U.S. developer space for open models, the company will have access to a mass of users it can drive to its chips and standards.
There are also growing questions about the cost of AI inference, which has companies exploring cheaper models built by Chinese companies like Moonshot, DeepSeek, and Alibaba. Right now, adoption is relatively small but growing — just 6% of companies use open-weight models, according to a survey of spending data by Ramp, or just 2% of software engineers measured by Jellyfish, which makes tools for developers.
Nik Albarran, the AI product lead at Jellyfish, told TechCrunch that open-weight models are primarily used by companies whose products rely on repeated inference workloads, like those providing customer service chats. Because these are high-volume tasks with a lot of repetition, an open-weight model can be tuned to answer the questions cheaply.
That’s certainly how Stripe has framed its OpenRouter acquisition. “Tokens are the central currency for companies building with AI, and it’s clear that the real-world economic potential will depend on making good use of scarce compute resources,” Patrick Collison, Stripe’s co-founder and CEO, said in a statement.
For coding and agentic tasks, however, varying requests and more reasoning mean that frontier models often win out, in part because the proprietary labs provide easier access, and in some cases a token subsidy. Albarran says that as companies dial in AI workflows, it will be easier to turn to open models. Still, the main reason companies look to those models now is for control and configurability, not because of spending concerns.
“There are not many companies where that is the case yet … [but] if the prices continue to go up from the frontier labs, more and more companies will be forced to at least consider it,” Albarran told TechCrunch. “When your AI-driven workflows are much more mature, that’s when it makes sense to invest in self-hosting models.”
Lin Qiao is the CEO of Fireworks, a leading open-weight models router and host for corporate users that is often discussed as a potential acquisition for a tech giant. Qiao says her company processes 40 trillion tokens a day, more than either of Gemini’s or OpenAI’s APIs.
Fireworks’ bet is on model diversity: As LLMs proliferate and improve, it will be easier for companies to train them specifically for their needs. “Every single app company should consider hiring an in-house researcher,” she told TechCrunch last week. “They can use their product and product data to build their own model. The future is actually specialized intelligence. Literally, every single company should have their own model per use case, and that will happen automatically.”
It’s easy to forget how early we are in the development of AI as a tool and a business. The dominance of OpenAI and Anthropic, however, isn’t inevitable. As the tech giants look to hedge their bets on the biggest labs, the allure of open technology is proving tough to resist.
