 |
|
|
|||||
Can a headline-making squabble with a client actually be good for a brand? This weeks dispute between the Department of Defense and Anthropic, a high-profile player in the super-competitive field of artificial intelligence, may be just that. The dispute involves whether the Pentagon, which has an agreement to use Anthropic technology, can apply it in a wider range of scenarios: all lawful use cases. Anthropic has resisted signing off on some potential scenarios, and the Pentagon has essentially accused it of being overly cautious. As it happens, that assessment basically aligns with Anthropics efforts (most recently via Super Bowl ads aimed squarely at prominent rival OpenAI) to burnish a reputation as a thoughtful and considered AI innovator. At a moment when the pros-vs.-cons implications and potential consequences of AI are more hotly debated than ever, Anthropics public image tries to straddle the divide. Presumably Anthropic (best known to consumers for its AI chat tool Claude) would prefer to push that reputation without alienating a lucrative client. But the underlying feud concerns how the military can use Anthropics technology, with the company reportedly seeking limits on applications involving mass surveillance and autonomous weapons. A Pentagon spokesman told Fast Company that the militarys relationship with Anthropic is being reviewed, adding: Our nation requires that our partners be willing to help our warfighters win in any fight. The department has reportedly threatened to label Anthropic a supply chain risk, lumping it in with supposedly woke tech companies, causing potential problems not just for Anthropic but for partners like Palintir. So far Anthropics basic stance amounts to: This is a uniquely potent technology whose eventualities we dont fully comprehend, so there are limits to uses well currently permit. Put more bluntly: We are not reckless. Not moving so fast that you break important thingslike user trust, or civilizationis a message thats of a piece with the official image Anthropic has sought to cultivate. The company was founded by OpenAI refugees who argued back in 2021 that the company was prioritizing monetization over safety. Its recent Super Bowl ads are the highest-profile example of this branding so far: directly mocking OpenAI for experimenting with advertising on its consumer-facing product ChatGPT, and presenting the results as a slop-dystopian mess. The spots were, as Fast Companys Jeff Beer explained, a rare example of straight-up ire slung at a category competitor. They could arguably be the first salvo in a branding battle akin to Apple vs. Microsoft, with Anthropic seizing the role of righteous challenger. (OpenAIs initial response included belittling Anthropics business, which just lends to the latters underdog pose.) As a brand image to shoot for, being the responsible AI player is an understandable goal. The technology has been divisive for years at this point, and lately thats reached a crescendo. Seen by many as a threat to privacy, a job-killer, an environmental menace, and a source of endless misinformation and slop, its simultaneously touted by Silicon Valley elites and their intellectual brethren as an unprecedented boon to humanity. The only point of agreement is that the changes will be big and fast and pretty much unstoppable. And no matter how much you already believe that, there is some guy on X arguing that you still dont really get it. No wonder there seems to be room for an AI company with a cautious message. Of course this is branding were talking about, and ultimately Anthropic is under the same marketplace pressures as its rivals. And its actual behavior hasnt always been pristine. Notably it agreed last year to pay a record $1.5 billion to settle a class-action lawsuit alleging its models trained on some 500,000 copyrighted books. Despite its Pentagon dispute, its technology is already intertwined with the American military, and was reportedly used in the recent U.S. capture of Venezuelan strongman Nicolás Maduro. And of course it may yet acquiesce to Pentagon demands. (According to Axios, Anthropics annual revenue is around $14 billion, and its Department of Defense deal is pegged at $200 millionnot chump change, but not existential.) Still, the squabble is an occasion for Anthropic to demonstrate that its rhetoric and actions line up. At the very least, that could be good for its flagship chat tool Claude: Consumers tempted by AI hype but worried about its potential downsides may see Anthropic as the fledgling technologys least-reckless major player. And given how divisive the AI category has become, that might count as a brand win.
Category:
E-Commerce
“I really want to see a mass driver on the moon that is shooting AI satellites into deep space,” Elon Musk said last week when he announced his plan to go to the moon. “It’s going to be incredibly exciting to see it happen.” He’s right. I want to see it too, although probably we will both be dead before his vision is realized. The lunar mass driveressentially a cannon that uses magnetic power to accelerate an objectis a key component to launch the million satellites Musk wants to put in orbit around the Earth. But Musk wasn’t the first person to come up with the idea. Smarter people than him thought about this in the 1970s as the solution to a key problem for human exploration. Launching spacecraft from Earth is extremely expensive. Every pound lifted from Cape Canaveral to low Earth orbit costs thousands of dollars in fuel, hardware, and operational complexity. The farther you want to go in space, the more massive and complex the rocket has to be, increasing costs. Chemical rockets must carry their own oxidizer and propellant, which means most of the vehicle’s mass is just fuel to lift more fuel. This tyranny of the rocket equation has strangled space development for seven decades, only slightly eased by the economics of reusable rockets like the Falcon 9. A mass driver could break that stranglehold by using electricity instead of explosives, turning launches into a utility-scale operation rather than a high-wire act. On the Moon, where gravity is one-sixth of Earth’s and there’s no atmosphere to create drag, this technology could launch payloads at a fraction of the costa few dollars per pound in electricity. Compare that to the $1,200 per pound it currently costs to launch a payload on a reusable Falcon 9 rocket. [Image: Xai] An elegant design American physicist Gerard O’Neill and MIT physicist Henry Kolm built the first prototype of a mass driver in 1976 with a $2,000 budget. The Mass Drive 1 could fire objects at 131 feet per second while experiencing 33 times Earth’s gravity. Their next version achieved 10 times greater acceleration with a comparable funding increase. University of Texas researchers subsequently priced a serious version at $47 million for a device capable of launching a 22-pound payload at 13,400 miles per hour. A mass driver is basically a very long track stretching across the lunar surface, angled gently skyward at its far end. The track is lined from end to end with hundreds of electromagnetic coils, which are simply loops of wire that snap into powerful magnets the instant electricity runs through them. A payload sits inside a magnetizable carrier called a bucket. To move the bucket, the coils fire in a precise sequence, one after another, each energizing at exactly the right moment as the bucket reaches it, grabbing it forward, then cutting off the instant it passes. The result is a cascade of invisible magnetic hands, each passing the bucket to the next. The bucket never makes mechanical contact with any surface: It is held aloft and guided entirely by the interplay of magnetic fields, which is why these systems have a theoretical operational lifespan of up to millions of launches with negligible wear. Musk describes it as a large maglev train, the same levitation technology that holds high-speed trains above their rails in Japan or China. But the mass driver reaches much faster speeds than any train on Earth: about 1.5 miles per second, enough to escape the gravity pull of the Moon. To achieve that speed, the mass driver uses two distinct engineered stages. In the first, the coils sit at equal intervals and their electrical timing locks to the bucket’s exact positioneach successive push arrives at precisely the right instant, so the force compounds as velocity builds. In the second stage, the interval between coils progressively widens, which paces the pushes further apart in distance and holds the rate of acceleration constant rather than letting it keep climbing so the increase in acceleration doesnt destroy the bucket or its cargo. At the terminal end of the track, the bucket releases its payloada xAI satellite according to Musk’s visioninto space at a minimum speed of 5,300 mph, enough escape the Moon’s gravity. The trajectory that the load follows depends on the position of the Moon at the moment of the launch, following the orientation of the mass driver relative to the space. Then the bucket gets caught by a braking system, recovered, and sent back to the beginning for the next launch. No combustion. No exhaust. No rocket equation. No problems. Its a beautiful solution. Its also doable, as ONeill and Kolm demonstrated practically. According to independent researcher and author Keith Sadlocha, a working lunar mass driver would require a track between 1,620 and 5,350 feet long, operating at accelerations between 30 and 400 times Earth’s gravity in standard operation. At those forces, only rugged, non-human cargo survives the ridewhich is exactly what Musk is planning. Musk has his sights set on manufacturing AI computing satellites on the lunar surface. The system can fire one payload every 10 to 11 seconds. Scaled to Musk’s stated target of one million satellites in orbit, that cadence, sustained continuously, is what makes the economics viable in a way no rocket ever could. But to accomplish this, you will need a lot of electricity. For Musk’s purposes, system requires 8.7 to 20 megawatts of continuous power, enough to run a small town. Delivering that on the lunar surface requires between 400,000 and 634,000 square feet of solar arrayssomewhere between seven and 11 NFL football fields’ worth of panels according to Sadlocha’s calculations and NASA’s estimates. That’s using solar panel’s with an efficieny of roughly 30%, the figure NASA uses for cells especially designed for space use. Since the Moon endures two weeks of total darkness every month, this means the mass driver either sits idle for half of every lunar cycle or relies on supplemental power to keep firing through the night. NASA and the U.S. Department of Energy are developing a solution: The Fission Surface Power (FSP) project, which builds on the earlier Kilopower research program to produce compact nuclear fission reactors targeting 10 to 40 kilowatts of continuous output each, capable of running for a decade without refueling. Each FSP reactor will produce enough electricity to power just a few homes. Bridging the full gap between those modest reactors and a mass driver that demands the output of a small power plant would require deploying them not one or two at a time, but in the hundreds. That is not a technology problem so much as a logistics oneevery reactor has to be launched from Earth, landed softly on the Moon, and connected to the grid before the mass driver fires its first payload. The program, however, is still in development and a lunar deployment is not expected before the late 2020s at the earliest. And thats extremely optimistic, given the constant delays of those nuclear projects and the Moon return plans. [Image: Xai] Unrealistic timeline Scaling from the $47 million 22-pound-launch prototype that University of Texas researchers projected to a working lunar installation capable of launching huge satellites is where you begin to feel just how vast the distance is between a compelling idea and a functioning machine. Sadlocha estimates that a full lunar mass driver system requires approximately 362 metric tons of hardware. Thats 24 heavy-lift rocket launches worth of components that must be manufactured on Earth, survive a 239,000-mile journey through the void, and be assembled by people wearing pressurized suits on a surface thatbathed by radiationis extremely hostile to humans. That surface is coated in lunar dust too, ultra-fine abrasive shaped by billions of years of micrometeoroid impacts into particles with microscopic cutting edges that never dulled, because there has never been wind or rain or any erosive force on the Moon to round them off. It clings electrostatically to visors, suits, seals, and coil windings alike. The payload carriers themselves face thermal melting at extreme velocities, demanding materials that do not yet exist in proven lunar-rated form. You can argue that maybe Musks Optimus robots can avoid this, but his robots can barely function on Earth. Musk’s stated plan is to mine lunar silicon and oxygen and manufacture the server hardware on the surfacea bootstrapping strategy that, if it works, would reduce Earth-launch dependency over time toward what he called a self-growing city capable of rapid expansion from local resources. The Moon does contain silicon, oxygen, helium-3, and water ice at the poles. But the superconducting coils at the heart of the mass driver require precisely manufactured materials that the lunar industry will not be able to produce in decades. Every critical component rides to the Moon on Starship until that changes. We know that Starship is so behind schedule that it has pushed the first mission back to the Moon from 2027s NASA projected time to 2028. Sadlocha rates the technology at readiness level 5 on NASA’s 1-to-9 scale components validated in laboratories, not yet tested in space. Realistic deployment, his study concludes, will take between 5 and 15 years from the moment serious investment begins. That can take the project into the 2040s, easily. Thats why Lluc Palerm, satellite research director at Analysys Mason, said to PC Magazine that Musk’s lunar server plan carries a magnitude of challenge equivalent to a Mars mission. But like we already pointed out, Musk’s timeline is fantasy, or “aspirational” as he qualifies his predictions. The gap between his ambitious renderings and actual functioning hardware remains a dream measured in decades, not the 10 years he’s promising investors before his planned June 2026 initial public offering targeting a $1.5 trillion valuation. Musk is no JFK, and building factories and a mass driver on the Moon is orders of magnitude more complex than just putting boots on the Moon like the Apollo program did. It’s doable, yes. We’ll get the mass driver, eventually. Just not on Musk time.
Category:
E-Commerce
The retail platform eBay is set to acquire fashion resale app Depop from Etsy in a $1.2 billion transaction. Ostensibly, the deal will help eBay to cultivate a new audience of Gen Z and Gen Alpha shoppers. But I think theres a deeper reason that eBay might want to lock Depop down: its simply the best looking resale interface out there right now. The deal was announced on February 18 in a press release from Etsy. Its expected to close some time in the second quarter of 2026, and, per an email sent to Depops customers, after the merger Depop will remain a stand-alone brand within eBay and retain its name, brand, and platform. For eBay, acquiring Depop makes a good measure of intuitive sense. Generally, resale is trending upward: Based on ThredUps 2025 Resale Report, the secondhand apparel market is expected to reach $367 billion by 2029, growing 2.7 times faster than the overall global apparel market. Millennials, Gen Z, and Gen Alpha shoppers are some of the strongest drivers of that trend, with 39% of younger generation shoppers having made a secondhand apparel purchase on a social commerce platform in the 12 months before the study was published. Depop is one of the top platforms for young people looking to buy and sell clothes. In 2025, the brand achieved approximately $1 billion in sales, including nearly 60% year-over-year growth in the U.S. As of December 31, it had seven million active customers, nearly 90% of which were younger than 34. That user base will be a major boon for eBay, who says that millennial and Gen Z consumers have been two of the biggest drivers of active buyer growth in the past three years. As a Gen Z vintage clothing enthusiast, Ive shopped on pretty much every resale site you can think of, from Poshmark and ThredUp to eBay, Facebook Marketplace, Mercari, and Etsy. Among all of these options, Depop is far and away the best resale site to look at and the easiest one to use. Thats not to say that Depop doesnt have any issuesa brief glance at the sites subreddit will reveal plenty of user grievances, not least of which is the tendency of certain Depop sellers to price a Brandy Melville baby tee at a cost that could put your checking account in the red. But from a pure UX and design standpoint, Depop is far outperforming its competitors by taking its major design cues from popular social media apps. And for a digitally native generation thats used to doing most of their shopping online, that makes a big difference. A social media-esque app experience Depop knows that its customers are young, tech savvy, and probably spending most of their phone time on social media sites like TikTok and Instagramand it shows in the apps design. Depop homepage [Screenshot: courtesy of the author] When you open the Depop app, youre immediately greeted with a Suggested for you page that’s functionally similarly to TikToks Explore feature. Here, the Depop algorithm presents you with an endlessly scrollable page of items curated based on your searches, likes, and saveslike how TikTok or Instagram might serve you videos according to your interests. In contrast, eBays app homepage looks more similar to a standard e-commerce webpage, directing users to its different goods categories and promoting whatever deals and sales are currently trending. eBay homepage [Screenshot: courtesy of the author] In terms of its layout, Depops app is simple and aesthetically pleasing. Its interface is almost identical to Pinterestwhere many customers are likely looking for outfit inspirationwith a subtle bar of categories at the bottom of the page and about four spotlighted items on view at a time. The minimalist information density encourages you to keep scrolling to find more items, rather than overwhelming you with a sea of information. Other apps, like Poshmark, eBay, and ThredUp, seem to opt instead for presenting users with a wealth of options to choose from when they first log on, which, counterintuitively, can make scrolling feel less appealing. While I tend to open my computer if I want to browse other retail sites, I almost always open Depop on my phone, and imagine it in a similar category to social media apps. Considering that the secondhand apparel market is becoming so popular among younger shoppers, other resale platforms might want to take note. Simple selling UI Over the past few months, Ive developed the niche hobby of restoring and selling vintage wedding dresses online (typically for a profit of about $10 apiece, but Im in it for the fun of the game). Having used both Depop and Etsy to sell my own products, I find Depops seler UX more intuitive and simple to follow. [Screenshot: courtesy of the author] From an app standpoint, Etsy has two separate platforms: one app for selling, and one app for buying. On its website, sellers also have to navigate to a shop management platform to look at their listings. Depop, on the other hand, is consolidated into one app experience, where sellers can manage all of their listings and their purchases. Creating an actual Depop listing feels akin to making a post on Instagram. Sellers navigate to a + at the bottom of the app (like Instagram), add a series of photos and a caption (also like Instagram), and include a few key tags for their item. Depop also handles shipping through a system that lets users provide an estimate of their items weight and then creates an appropriate label. Shipping on Etsy is more seller-directed. While some more experienced sellers might prefer Etsys approach, Depops feels more beginner-friendly. A final refuge from AI listings One of my biggest personal gripes with the current state of resale shopping is the absolute deluge of AI-generated product images that seem to have flooded certain sites over the past few months. From top: Depop, eBay, Etsy. [Screenshot: courtesy of the author] In my experience, eBay and Etsy are the biggest offenders of this trend. Searching for the term fantasy dress, on eBay and Etsy, for example, leads to at least one out of four top results with all the hallmarks of an AI image. The same search on Depop yields results that all seem to be real photographs. This example is just one small microcosm of the shopping experience on these sites: while AI photos are becoming increasingly common in resale, buying on Depop still largely feels like sifting through a strangers closet, which was the sites original charm. Its unclear exactly why AI photos seem less prominent on Depop; though it may be related to the companys regulations against stock photos. In its guidelines, Depop instructs sellers to Only use photos taken by yourself. Depop didnt immediately respond to Fast Companys request for comment on its AI imagery policies.
Category:
E-Commerce
All news |
||||||||||||||||||
|
||||||||||||||||||