Welcome back to TechCrunch Mobility — your central hub for news and insights on the future of transportation. Sign up here for free — just click TechCrunch Mobility! About 39.4 million of those folks will use a car. This edition has news on loads of companies, including Aurora, Uber, Tesla, and Waymo. Plus, a number of startups you may be interested in. Luminar, the lidar startup turned SPAC, appears to be grasping for capital. You might recall Luminar's board recently replaced founder Austin Russell as its CEO. The company is also going through another restructuring — its third in a year. SparkCharge, which offers what it calls “charging-as-a-service” for fleets, raised $15.5 million in a Series A-1 round led by Monte's Fam, with participation from Cleveland Avenue, Collab Capital, Elemental Impact, MarcyPen, and non sibi ventures. Alongside the equity round, SparkCharge also secured a $15 million venture loan from Horizon Technology Finance Corp. Sylndr, a Cairo-based online used car sales startup that is expanding into auto financing, servicing, and tools for dealers, raised $15.7 million. The round was led by Development Partners International's Nclude Fund. Aurora has put human “observers” in its self-driving trucks at the request of its partner PACCAR, a disclosure that has some scratching their heads about the move. To be clear, these “observers” are not human safety operators, meaning they can't intervene. This news prompted folks to send me a slew of messages with questions like “Why?” and “What's the point?” Einride founder Robert Falck is stepping down from the role of CEO. Einride's CFO, Roozbeh Charli, will take over the role of chief executive effective immediately. Reliable Robotics, the autonomous aviation company, appointed Marc Stoll as its new CFO. Zoox has completed the “initial mapping phase” and will begin testing its self-driving vehicles in Atlanta later this summer. The California Public Utilities Commission approved Waymo's request to expand its commercial robotaxi service area into more communities south of San Francisco. Meanwhile, Waymo and Uber plan to start offering robotaxi rides in Atlanta to select customers who signed onto a waitlist earlier this year. Using a geofence represents a major strategy shift for Musk, who spent years claiming his company would be able to create a general-purpose self-driving solution that could be dropped into any location and work without human supervision. That includes an expansion of Insights AI, which Uber Freight quietly launched in 2023, as well as more than 30 AI agents built to “execute key logistics tasks throughout the freight lifecycle.” Senior reporter Sean O'Kane interviewed CEO Lior Ron about the company's dive into AI — including the how, why, and what's next. Yup, “This week's wheels” is back with Rebecca Bellan, who writes about her time on the new Heybike Alpha, a sturdy, fat-tire, all-terrain e-bike with a $1,699 price tag. But generally, Rebecca felt the Alpha was an excellent all-around e-bike, whether you want to take it on off-road adventures or use it in the city to do your weekly Trader Joe's shopping. TechCrunch Mobility: Uber Freight's AI bet, Tesla's robotaxi caveat, and Nikola's trucks hit the auction block OpenAI upgrades the AI model powering its Operator agent Microsoft says its Aurora AI can accurately predict air quality, typhoons, and more Discord seeks to solve a problem that it created OpenAI goes all in with Jony Ive as Google plays AI catchup
Looking on were more than 200 people who had been invited to the Trump National Golf Club for a private gala dinner. Four of the guests agreed to tell WIRED about their experience. By comparison to Trump's previous banquets, thronging with D.C. insiders and members of the Silicon Valley elite, the crypto dinner attracted a mismatched collection of oddballs: independent traders rubbed shoulders with crypto executives, diehard Trump fans and even professional sportspeople—former NBA player Lamar Odom towered overhead. A handful wore bowties in Bitcoin orange; others sported gold Trump sneakers. Just after 7 pm, the dinner guests gathered at the window to watch Trump descend in Marine One, his presidential helicopter. A short while later, he appeared from behind a blue velvet curtain to whoops and applause from the crowd. From behind a lectern at one end of the dining room, backdropped by four US flags, Trump delivered a characteristically winding and digressive speech that sources say lasted around 25 minutes. At some point, he got round to crypto. A lot of people are starting to believe in it … This is really something that may be special—who knows, right? When Trump first promoted his memecoin in January, three days before the inauguration, the limited amount released into circulation rose in value to $14 billion. The remaining 80 percent of the supply is controlled by CIC Digital LLC—a subsidiary of a conglomerate owned by the Trump family—and Fight Fight Fight LLC, formed by long-time Trump ally Bill Zanker. With little more than a social media post, Trump had added billions of dollars to his paper net worth. The dinner competition revived objections among critics who feared that the memecoin could be used as a vector for bribery. Theoretically, by making a large investment in TRUMP, driving up the price, critics worried that a politically-motivated actor could discreetly curry favor with the president. The dinner, critics argued, compounded that risk by making the unsavory arrangement explicit: a large investment in return for an encounter with Trump. “What's happening tonight at Trump's golf course is, in effect, putting a ‘for sale' sign on the White House. It's auctioning off access,” claimed Democrat senator Richard Blumenthal on Thursday morning at a press conference hosted by non-profit Accountable.US. The China-born entrepreneur has become increasingly entangled in the Trump family's expanding web of crypto ventures; in addition to investing in the memecoin, Sun previously disclosed a combined $75 million investment in a separate crypto coin issued by World Liberty Financial, a company affiliated with the Trump family and promoted by Eric and Donald Trump Jr. In early May, Eric Trump announced a partnership between World Liberty Financial and TRON, a crypto network developed by Sun. Under the Biden administration, US regulators had charged Sun and several of his companies with market manipulation and offering unregistered securities. In February, after Trump had returned to the White House, a district judge granted a joint request to stay the case in order for both sides to negotiate an out-of-court resolution. The event organizer did not respond immediately to a request for comment. President Trump only acts in the best interests of the American public, which is why they overwhelmingly re-elected him to this office, despite years of lies and false accusations against him and his businesses from the fake news media.” “You don't get to meet the president easily,” Vincent Liu, chief investment officer at trading firm Kronos Research, told WIRED a few days before the dinner. Others hoped to run into famous members of Trump's inner circle—perhaps Elon Musk or David Sacks, the venture capitalist now serving as crypto and AI czar to the White House. “If it's someone from the PayPal Mafia, I'll probably tell them that Peter Thiel's book changed my life,” said 25-year-old TikTok prankster Nicholas Pinto, who purchased around $300,000 in TRUMP to earn his seat at the dinner. To celebrate winning a place at the dinner, Pinto had spraypainted “Hold $Trump” onto the chassis of his blue Mercedes G-Wagon. One independent crypto trader, who spoke to WIRED on condition of anonymity, didn't really intend to win a place at the dinner at all. They bought into TRUMP with the aim of profiting by a potential uplift in price brought about by the competition, but currently faced with a loss on their trade if they were to sell, instead chose to settle for a lavish dinner. Some signed their respective leaderboard usernames in marker pen. Others made their way towards circular tables, each seating 10 people, arrayed beneath a set of crystal chandeliers. Waiting on the chairs were gift bags containing Fight Fight Fight-themed hats and posters, and a collectible plastic card (some allege that they didn't receive merch at their seats.) The four largest coin holders—along with two other attendees selected by raffle, sources say—received a gem-encrusted Trump gold watch. Between mouthfuls, the attendees discussed trading and investment strategies—and Trump's speech. But others complained about the brevity of Trump's appearance: After his speech, Trump had departed immediately in a golf cart bound for his helicopter. “Trump could have at least given the top people their watches himself,” says Pinto. “It was the worst food I've ever had at a Trump golf course,” says Pinto, who added he left hungry. “I appreciate like everything the Trump administration done for our industry [sic],” said Sun. Big Story: The worm that no computer scientist can crack Yuval Noah Harari: “Prepare to share the planet with AI superintelligence
It's a misconception that transformers reason in token space. The final layer of a decoder only transformer has full access to entire latent space of all previous latents, the same latents you can project into a distribution of next tokens. That's essentially the core idea in Coconut[1][2], to keep the reasoning traces in a continuous space. Reasoning is definitely not happening in the linear projection to logits if that's what you mean. Apparently the reasoning doesn't even need to be correct during training?I guess it suggests to me that the reason CoT helps is that the model gets more compute to think internally, not that the words it produces are meaningful. I'm surprised nobody has come up with a good scheme for adaptive compute per token yet. I guess it suggests to me that the reason CoT helps is that the model gets more compute to think internally, not that the words it produces are meaningful. I'm surprised nobody has come up with a good scheme for adaptive compute per token yet. How do we know if the reasoning was correct or not? Do we have more information about what the model was thinking besides just what it says it was thinking? It's no different than how in English we can signal that a statement is related to a kind of politics or that it's about sex through particular word and phrase choice.Training for reasoning should be expected to amplify the subtext, since any random noise in the selection that by chance is correlated with the right results will get amplified.Perhaps you could try to dampen this by training two distinct models for a while, then swap their reasoning for a while before going back-- but sadly distinct models may still end up with similar subtexts due to correlations in their training data. Maybe ones with very distinct tokenization would be less likely to do so. Training for reasoning should be expected to amplify the subtext, since any random noise in the selection that by chance is correlated with the right results will get amplified.Perhaps you could try to dampen this by training two distinct models for a while, then swap their reasoning for a while before going back-- but sadly distinct models may still end up with similar subtexts due to correlations in their training data. Maybe ones with very distinct tokenization would be less likely to do so. Perhaps you could try to dampen this by training two distinct models for a while, then swap their reasoning for a while before going back-- but sadly distinct models may still end up with similar subtexts due to correlations in their training data. Maybe ones with very distinct tokenization would be less likely to do so. I hope that research into understanding LLM qualia eventually allow us to understand e.g. what it's like to [be a bat](https://en.wikipedia.org/wiki/What_Is_It_Like_to_Be_a_Bat%3F) We have our own personal 'culture' too-- it's just less obvious because its tied up with our own hidden state. that are absolutely not explicitly in the text immediately come back to you, stuff that no one or maybe only a spouse or very close friend might think.I think it may be very hard to explore hidden subtext because the signals may be almost arbitrarily weak and context dependent. The bare model may need only a little nudge to get to the right answer and the you have this big wall of "reasoning" where each token could carry very small amounts of subtext that cumulatively add up to a lot and push things in the right direction. I think it may be very hard to explore hidden subtext because the signals may be almost arbitrarily weak and context dependent. The bare model may need only a little nudge to get to the right answer and the you have this big wall of "reasoning" where each token could carry very small amounts of subtext that cumulatively add up to a lot and push things in the right direction. An example of this is toki pona, a minimalist constructed human language that is designed to only express "positive thinking".
Elon Musk has purportedly stepped away from his government duties. Courts are trying to strike down some of DOGE's most egregious efforts. It may seem as though the worst excesses of DOGE have passed, replaced by something closer to a stasis. While the image of DOGE most likely burned into your retina is that of Elon Musk wielding a literal chain saw, the theatrics belie an organization that has quietly permeated all corners of the federal government. Removing DOGE at this point would be like trying to remove a drop of food coloring from a glass of water. Giving the Justice Department an assist on alleged voter fraud indictments. Even if Musk claims he's stepping away—though he met with House Republicans just this week—his lieutenants are still firmly in place at the agencies that control the federal workforce and regulate his companies. DOGE is reportedly using his xAI Grok chatbot to parse sensitive data, which potentially means millions of Americans' personal information is doubling as training data for the model. The Trump administration used it as a battering ram to push its policies through with overwhelming force. While the courts have held firm in some cases—just this week, a judge declared DOGE's takeover of the United States Institute of Peace to be unlawful—the policies still exist, and there are other ways to achieve them. He has been explaining for months what happens after DOGE's first assault. “We're going to use all of our executive tools to make those savings permanent,” Vought said in an interview with Fox Business anchor Larry Kudlow on March 11. The only part of this that hasn't gone to plan is that Congress may be more amenable than Vought gave it credit for; the House of Representatives managed to pass Donald Trump's One Big Beautiful Bill this week, complete with provisions that cut the social safety net into ribbons. In some ways, DOGE is even more dangerous when it's boring. It has always been a tool, not an engine unto itself. It's no longer useful to think of it as a separate entity, as a tech billionaire's personal strike force. It's a means to an end, one part of a larger project to gut the federal government and redefine the social contract. It's important to keep a close eye on DOGE, to continue shining a light on what it does in the dark. What do you think comes next with DOGE and Elon Musk? Leave a comment on the site or send your thoughts to mail@wired.com. • DOGE Used a Meta AI Model to Review Emails From Federal Workers: Materials viewed by WIRED show that DOGE affiliates within the Office of Personnel Management tested and used Meta's Llama 2 model to review responses to the infamous “Fork in the Road” email. • FEMA Has Canceled Its 4-Year Strategic Plan Ahead of Hurricane Season: Multiple FEMA employees tell WIRED that they did not know of another time when a strategic plan was rescinded without another in place. • What It's Like to Interview for a Job at DOGE: WIRED spoke with someone who applied for a job at Elon Musk's so-called DOGE and discussed the five-step hiring process. 🔗 Trump Administration Says It Is Halting Harvard's Ability to Enroll International Students: This is a massive escalation in the Trump administration's assault on universities. 🔗 Musk's DOGE Expanding His Grok AI in US Government, Raising Conflict Concerns: DOGE will apparently soon be using Elon Musk's AI chatbot Grok to analyze federal government data. 🔗 The Case of the ‘Lost' FOIA Requests: A few months ago, a number of FOIA requests at federal agencies were lost. The data was compromised or deleted by two convicted hackers. This week on our flagship Uncanny Valley podcast, Katie Drummond, Lauren Goode, and Michael Calore debated whether using AI in school is actually cheating. Big Story: The worm that no computer scientist can crack Yuval Noah Harari: “Prepare to share the planet with AI superintelligence” WIRED may earn a portion of sales from products that are purchased through our site as part of our Affiliate Partnerships with retailers. The material on this site may not be reproduced, distributed, transmitted, cached or otherwise used, except with the prior written permission of Condé Nast.
This article is part of Gizmodo Deals, produced separately from the editorial team. We may earn a commission when you buy through links on the site. If you have been looking to streamline your workspace and take it to the next level with a single, powerful hub, then the Anker Prime Docking Station, 14-Port with 160W Max Output is the perfect fit for you. Whether you're managing multiple displays, transferring large files, or charging devices, this docking station is designed to handle it all fast and efficiently. For a limited time only, you can acquire the Anker Prime Docking Station, 14-Port with 160W Max Output for $189.99, down from its original retail price of $269.99. This deal is ideal for those seeking a solution that you just won't find anywhere else. This docking station has a total output of 160W, capable of quickly charging up to four devices simultaneously through its USB-C and USB-A ports. That's all the essentials you might need before leaving your house, including your phone, being charged at the same time without sacrificing any power. You will be able to transfer files from an external device to your computer seamlessly while also doing something else on another display monitor. The integrated GaN power module not only delivers high power, but also maintains a compact and space-efficient design, making it a perfect fit for basically any desk setup. Beyond all this impressive functionality, the Anker Prime Docking Station also features a modern design that complements any workspace. Its vertical orientation saves valuable desk space, while the professional finish and minimalist aesthetic add a touch of sophistication to the whole thing. The real-time smart interface provides immediate visual feedback on power levels and connection statuses, ensuring you are always informed about your setup and the status of your devices. This combination of style and practicality makes it not just a powerful tool, but an enhancement to your overall work environment. At only $189.99, the Anker Prime 14-in-1 Docking Station offers exceptional value for professionals and tech enthusiasts alike. Get the best tech, science, and culture news in your inbox daily. News from the future, delivered to your present. We may earn a commission when you buy through links on our sites.
“As we have said before, this reckless misuse of the Congressional Review Act is unlawful, and California will not stand idly by.” Other Senate votes Thursday repealed waivers that allowed California to set stricter emissions standards for medium- and heavy-duty vehicles. Beginning in 2026, the state was to begin requiring increased sales of zero-emissions cars and passenger trucks until 2035, when automakers would have to sell only zero-emissions vehicles. Currently, two technologies qualify: hydrogen fuel cells and battery electric vehicles. Given the growing pains that fuel cells and hydrogen filling networks have experienced, EVs quickly became the de facto approach to meeting California's 2035 deadline. “We don't think it's an appropriate use of the Congressional Review Act, and we're prepared to defend ourselves if it's wrongfully weaponized,” he told Politico in early March. Updated 7:37 a.m. Pacific: Added news of California preparing a lawsuit against the federal government. Microsoft says its Aurora AI can accurately predict air quality, typhoons, and more OpenAI goes all in with Jony Ive as Google plays AI catchup Digg founder Kevin Rose offers to buy Pocket from Mozilla ChatGPT: Everything you need to know about the AI-powered chatbot California prepares to sue feds after Senate revokes the state's EV rule
In the startup world, access to cutting-edge tools isn't the biggest obstacle — it's knowing how to wield them with precision. At TechCrunch Sessions: AI, taking place on June 5 at UC Berkeley's Zellerbach Hall, we're digging into the frameworks and decisions that determine whether an AI startup can scale — or stall. We're excited to welcome one of the most tactical voices in that conversation: Iliana Quinonez, Director of North America Startups Customer Engineering at Google Cloud. Quinonez leads a technical team that partners with high-potential startups across the U.S. and Canada, offering hands-on support as they integrate Google Cloud and AI tools from pre-seed through IPO. Startups building with AI agents today face a layered set of questions. What's the right architecture for scaling intelligent behavior? Where do APIs end and core IP begin? Quinonez will tackle these questions head-on in a session designed to help founders cut through noise and make defensible decisions about infrastructure, model orchestration, and collaboration. With over two decades of leadership experience at Salesforce, SAP, and BEA Systems, she's helped both enterprise giants and early-stage startups map product ambition to practical execution. Her team also partners closely with accelerators, VCs, and developer ecosystems — giving her a panoramic view of what's working in AI … and what's not. Expect a clear-eyed discussion on the risks and rewards of building with AI agents, the tools startups are relying on, and what it really takes to democratize access to advanced machine learning — without compromising on speed or security. This is just one of many conversations happening at TC Sessions: AI — where the focus isn't just on the future of AI, but also on how to build it right now. You'll hear from leaders at OpenAI, Anthropic, Cohere, and Google Cloud, in sessions that span everything from foundational model strategy to data stack design. ChatGPT: Everything you need to know about the AI-powered chatbot California prepares to sue feds after Senate revokes the state's EV rule X continues to suffer bugs following Thursday outage
This article is part of Gizmodo Deals, produced separately from the editorial team. We may earn a commission when you buy through links on the site. If you have been eyeing a new top high-end audio gear or have been considering upgrading your current one, there is no better time than now. Whether you're commuting, working, or simply relaxing, these amazing headphones have no match and provide an immersive listening experience that cannot be topped. That is a solid 34% discount on a pair of headphones that consistently receive high praise from users and critics all over the globe. This deal is part of Amazon's early Memorial Day sales, so it won't last long. The Bose QuietComfort Bluetooth Headphones boast up to 24 hours of continuous battery life, ensuring you can enjoy your favorite tunes all day and without any interruption. The over-ear design, combined with plush ear cushions, provides a comfortable fit for extended listening sessions. With built-in microphones and intuitive controls, you can easily manage both calls and music playback without even reaching your pocket for your device. The Bose QuietComfort headphones also feature advanced Bluetooth connectivity, allowing for seamless pairing with your devices. The active noise-canceling technology effectively blocks out ambient noise, making them ideal for travel, work, or any environment where you want to focus on your audio. This incredible headset, besides its amazing features, is also fashionable and incredibly comfortable to use. Not only will you be getting top-notch sound with an immersive sound quality that is unmatched, you will also be acquiring a piece of audio gear that will endure, due to its amazing materials, and with a battery life so potent that you will forget there are no wires between your device if choosing and the headset. At $229, the Bose QuietComfort Bluetooth Headphones offer exceptional value for those seeking high-quality audio and comfort. Visit Amazon now to secure your pair before the deal ends, or they run out of stock. Get the best tech, science, and culture news in your inbox daily. We may earn a commission when you buy through links on our sites.
This article is part of Gizmodo Deals, produced separately from the editorial team. We may earn a commission when you buy through links on the site. Even during a huge tech sale event like Memorial Day Weekend, this portable charger from Anker really should cost more than the mere $35 you can get it for at Best Buy. But if that's how low they're willing to go on the 20,000 mAh Anker Power Bank, which brings 30W of charge to your laptop, smartphone, or tablet, who are we to say no? The 6-inch length probably disqualifies it from being called pocket-sized, but it's a little under 3 inches wide and just an inch thick, and weighs in at just 1 pound. If you don't have a reliable charger in your travel bag for your summer trip plans, this is the one you want. It's also a safe charge for your devices, since the Power Bank is equipped with Anker's intelligent temperature monitoring system to prevent your devices from being damaged by overheating, keeping them at least 10 percent cooler than international safety standards, and also keep the Power Bank itself from becoming overly hot to the touch. The Anker Power Bank is also platform-agnostic — your Apple devices and MacBooks will charge just as readily as a Samsung smartphone or Galaxy Buds, Google devices, or most other leading brands. Another great perk — never losing your cord. The regular retail price of $46 was already a great deal for a compact 30W portable charger that's made to power up devices large and small and across brands and platforms, but this Best Buy offer dropping the price to just $35 makes this one of the better Memorial Day Weekend deals to cash in on. Head to Best Buy now to buy the Anker Power Bank in person, or order it online. Get the best tech, science, and culture news in your inbox daily. News from the future, delivered to your present. We may earn a commission when you buy through links on our sites.
All products featured on WIRED are independently selected by our editors. I've made no secret about my love for coffee makers from the Portland-based company Ratio, ever since it came out with the Ratio Eight model a decade ago. Two of WIRED's favorite drip coffee makers on earth are made by Ratio, in fact. As well as anybody, the Portland company has managed to take the finicky techniques of third-wave barista pour-over and merge them with automatic drip coffee to create some of the most full-bodied and aromatic cups of coffee I've had the pleasure of trying. The Ratio Four (8/10, WIRED Recommends) remains my favorite single-serve drip coffeemaker made by anyone, anywhere. But it all began a decade ago with the Ratio Eight, which helped change the conversation around what a home drip coffeemaker could be, adopting the aesthetics and a lot of the techniques surrounding barista café pour-over but applying them to an automated home coffee machine— using locally hewn wood and the shapely curves of a Chemex-style pour-over coffeemaker. The new 8-cup Ratio Eight Series 2, slated for release in September, is now available for presale on Ratio's website. For a limited time, preorders are discounted to $639. But by September, when the Eight Series 2 is projected to go on sale for real, that price will shoot up to $799. That's among the highest prices I'm aware of for a drip or pour-over coffee maker. I can say I got to try out the new Ratio Eight Series 2 at Ratio's offices this week, just before preorders launched. But the coffee I had was full-bodied, full-flavored, and delicious. If you're a fan of the original Ratio Eight, chances are you were drawn in first by the aesthetics. And that pretty hourglass shape of the blown-glass carafe, with the conical brewing filter best known from café pour-over. But Ratio has learned a few things since making the original Eight, said Ratio founder and CEO Mark Hellweg, at a meeting this month that also included the Eight's designer, James Owen. One of the things Ratio learned is that using a closed, flat-bottom brewing chamber leads to much more consistent coffee than conical glass. It's easier to control both in terms of water flow and brewing temperature, and is more forgiving if you don't have a super-expensive grinder. (That said, you know, we like expensive grinders and recommend a few.) Hellweg says they'll indeed keep making the Chemex-style carafe as an option. But the Series 2, by default, will look much different, though it still has its hourglass shape as a link to the past. Instead, the new brewing basket will make use of technology that Ratio developed when making two of my favorite coffeemakers, the Ratio Four and the Ratio Six Series 2, which consistently feel like magic tricks to me when I use them at home. Few others can manage such full-bodied, consistently thick extraction on automatic drip or pour-over-style coffee, while being so simple to use. First, there's the “bloom,” which is a poetic name for pre-infusing coffee grounds with a small amount of hot water. This releases the carbon dioxide trapped in fresh beans, which would otherwise interfere with extraction when the hot water begins flowing in earnest. Another thing that makes café pour-over coffee express flavor so well is agitation. Basically, if you drop water onto coffee, or shake it, or spin it, the coffee will release its secrets better. On the Eight Series 2, when Hellweg brewed a batch at Ratio's Portland headquarters, this meant that water extracted evenly across the entire brew basket, with no obvious channeling. So the falling water droplets extract both faster and evenly. Temperature control is a more interesting and complex conversation. Ratio is one of a small number of coffee-maker companies who've had their devices certified by the international Specialty Coffee Association as hewing to a very tight set of criteria, which includes holding temperatures within a tight 4-degree (Celsius) range during the entire brewing process. But even this doesn't tell the whole story. Ratio does a lot of sophisticated firmware rejiggering to play around with temperature curves during brewing—what chief operating officer Bradley Walhood, in an excitable moment, called “a literal crapload” of testing and firmware updates. Though the Ratio Eight Series 2 is a 1,400-watt device, they intentionally don't use all of it at all times, changing energy consumption in order to change how the heating element works over time and get the best-tasting brew. Like Ratio's previous Four, the Eight Series 2 will have two different brew modes for small and large batches of coffee, to further refine temperature curves and keep them in optimal ranges. There's a lot of interest in plastic-free everything, in these days when new and frightening revelations about microplastics seem to come once a week. But it's difficult, and very expensive, to make a truly plastic-free automatic coffee machine. Plastic-free brewing is a bit of a holy grail, achieved mostly by much older, lower-tech methods like French press, moka pot, and pour-over. The cold-water reservoir is a BPA-free Tritan copolymer. And there's a bit of BPA-free plastic used in the heating element that doesn't come into contact with hot water, said Hellweg. What all this means is that heated water or coffee never touches plastic. “Check this out,” said Hellweg, with evident pride, handing me a small and weighty piece of steel that could double as a stress fidget. It actually did feel good, the weight of this lid. Plastic makes you forget the satisfying solidity of cast metal. These more expensive materials are reflected, of course, in the price. The preorder window offers a quite substantial discount on the anticipated retail price, but it's still not low. They're also double the price of Ratio's own Six and Four. Will the coffee alone justify this higher price? It's hard to know based on one delicious carafe of coffee. That's the sort of thing I'll only learn over time, when testing side-by-side. The new thermal carafe, with its stainless steel construction, is sturdy and lovely. Ratio's owners and designers also cited instability surrounding tariffs, which is almost certainly a factor in announcing a price that'll first take effect six months from now. The Eight Series 2, like a large portion of products from small companies in the United States, is manufactured in China because American factories aren't set up for custom tooling on small production runs. And that based on past experience with Ratio, the Eight Series 2 will probably remain the best-tasting cup of drip coffee I can get by just pressing a single button. Exclusive: Up To 50% Off 6 Boxes With Factor Promo Code WIRED may earn a portion of sales from products that are purchased through our site as part of our Affiliate Partnerships with retailers. The material on this site may not be reproduced, distributed, transmitted, cached or otherwise used, except with the prior written permission of Condé Nast.
When you purchase through links on our site, we may earn an affiliate commission. There are dozens of solid-state drives with a PCIe 5.0 interface over at Computex 2025, and they no longer surprise us; however, we have seen a drive that features a PCIe 6.0 x4 interface and can potentially boast a 30.25 GB/s sequential read and write speed at the trade show. Micron's 9650 Pro SSD with a PCIe Gen6 interface will unlikely launch anytime soon, but for now, the unit is an important test vehicle for companies like Astera Labs, which plays an important role in developing next-gen AI platforms. At the show, Astera Labs used Micron's PCIe 6.0 SSD to demonstrate its Scorpio PCIe 6.0 4x16 switch, Aries Bandwidth-matching Gearbox software, and Aries 6 timers. For example, PCIe 6.0 switches can enable peer-to-peer communications between AI GPUs and SSDs, bypassing the CPU. In addition, when paired with Astera's Gearbox software and hardware, they can reduce the number of PCIe 6.0 lanes required to work with PCIe 5.0 hosts (e.g., to achieve a PCIe 5.0 x8 performance only for PCIe 6.0 lanes are needed) and therefore enable instalation of more drives into a box, which can be critical for some AI systems. Certification of PCIe 6.0 devices by PCI-SIG was delayed from mid-2024 to the second half of 2025, so while GPUs like Nvidia's Blackwell support the technology, they have not passed official interoperability tests. While EVT3 is still considered pre-production, by the third revision, most hardware issues should be resolved, and firmware/software should be approaching maturity. There are two iterations to go through: DVT (Design Validation Test) to validate the design under production conditions, and PVT (Production Validation Test) for full qualifications by customers. We do not know whether Micron will wait for the PCI-SIG to initiate interoperability tests in the second half of this year or will first start qualifications of its 9650 Pro PVT drive with select customers interested in PCIe 6.0 storage, but technically, EVT3 can be used for PCI-SIG's compatibility validation. Get Tom's Hardware's best news and in-depth reviews, straight to your inbox. Anton Shilov is a contributing writer at Tom's Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends. Tom's Hardware is part of Future US Inc, an international media group and leading digital publisher. © Future US, Inc. Full 7th Floor, 130 West 42nd Street, New York,
Taking away vaccines from people who might benefit: not so much. This week, two new leaders at the US Food and Drug Administration announced plans to limit access to covid vaccines, arguing that there is not much evidence to support the value of annual shots in healthy people. New vaccines will be made available only to the people who are most vulnerable—namely, those over 65 and others with conditions that make them more susceptible to severe disease. Covid vaccines will soon be required to go through more rigorous trials to ensure that they really are beneficial for people who aren't at high risk. No vaccine is perfect, but these medicines are still saving millions of lives. The plans have been met with fear and anger in some quarters. But they weren't all that shocking to me. In the UK, where I live, covid boosters have been offered only to vulnerable groups for a while now. Given that people respond differently to both the virus and the vaccine, perhaps individuals should be able to choose whether they get a vaccine or not. I should start by saying that covid vaccines have been a remarkable success story. The drugs were developed at record-breaking speed—they were given to people in clinical trials just 69 days after the virus had been identified. And they rescued many of us from lockdowns. But while many of us have benefited hugely from covid vaccinations in the past, there are questions over how useful continuing annual booster doses might be. That's the argument being made by FDA head Marty Makary and Vinay Prasad, director of the agency's Center for Biologics Evaluation and Research. He made incorrect assumptions about the coronavirus responsible for covid-19 and predicted that the disease would be “mostly gone” by April 2021. Prasad has said “the FDA is a failure” and has called annual covid boosters “a public health disaster the likes of which we've never seen before,” because of a perceived lack of clinical evidence to support their use. Makary and Prasad's plans, which were outlined in the New England Journal of Medicine on Tuesday, don't include such inflammatory language or unfounded claims, thankfully. In fact, they seem pretty measured: Annual covid booster shots will continue to be approved for vulnerable people but will have to be shown to benefit others before people outside the approved groups can access them. Let's address a few of the biggest ones. Each year, a flu vaccine is developed to protect against what scientists predict will be the dominant strain of virus circulating come flu season, which tends to run from October through March. “We seem to be getting waves of covid year-round,” she says. And an annual shot might not offer the best protection against covid anyway, says Fikadu Tafesse, an immunologist and virologist at Oregon Health & Science University in Portland. His own research suggests that leaving more than a year between booster doses could enhance their effectiveness. “One year is really a random time,” he says. It might be better to wait five or 10 years between doses instead, he adds. “If you are at risk [of a serious covid infection] you may actually need [a dose] every six months,” says Tafesse. “But for healthy individuals, it's a very different conversation.” There are reports that pediatricians are concerned about the impact on children, some of whom can develop serious cases of covid. “If we have safe and effective vaccines that prevent illness, we think they should be available,” James Campbell, vice chair of the committee on infectious diseases at the American Academy of Pediatrics, told STAT. Data show that older adults and people with underlying illness need the vaccine most. This question has been on my mind for a while. My two young children, who were born in the UK, have never been eligible for a covid vaccine in this country. My kids were eventually offered a vaccine in the US, when we temporarily moved there a couple of years ago. But by that point, the equation had changed. I was relieved to hear that Tafesse had made the same decision for his own children. The UK's Joint Committee on Vaccination and Immunology found that the benefits of vaccination are much smaller for children than they are for adults. “Of course there are children with health problems who should definitely have it,” says Dunachie. Research suggests that people who are vaccinated against covid-19 are less likely to end up transmitting the infection to the people around them. Having said that, the cost-benefit calculation will differ between individuals. For them, it might make total sense to get the vaccine. Dunachie thinks people should be able to make their own decisions. It is still not entirely clear who will be able to access covid vaccinations under the new plans, and how. Yes, we need up-to-date evidence to support the use of any drugs. But taking vaccines away from people who have experience with them and feel they could benefit from them doesn't feel like the ideal way to go about it. This article first appeared in The Checkup, MIT Technology Review's weekly biotech newsletter. To receive it in your inbox every Thursday, and read articles like this first, sign up here. Human “bodyoids” could reduce animal testing, improve drug development, and alleviate organ shortages. Pigs whose DNA makes them resistant to a virus could be the first big consumer product using gene editing. The multimillionaire longevity influencer thinks his new faith could save humanity from superintelligent AI. Colossal Biosciences claims it has revived an extinct species, but scientists outside the company are skeptical. Discover special offers, top stories, upcoming events, and more. Try refreshing this page and updating them one more time.