Composite 58Trend 10
The Polymarket betting pool wants to cash in on Hantavirus
Prediction markets have now turned their focus to hantavirus, a rare but severe category of viruses transmitted from rodents to humans, after several cases were identified earlier this month aboard an Atlantic cruise operated by Oceanwide Expeditions.
It’s a serious situation that has drawn global concern: Several passengers have tested positive for the illness, at least three cruise participants have died, and a number of others on the trip are reportedly experiencing symptoms.
Amid growing anxiety about the illness and, no doubt, memories of the nerve-racking first weeks of the Covid-19 pandemic, some people have taken to prediction markets to bet on what might happen next. On Polymarket, users have now invested about $3 million betting on whether there will be a Hantavirus outbreak this year. About $170,000 has been similarly wagered on Kalshi, as of late Friday afternoon. Both bets are set up to be resolved by the end of 2026, so there’s plenty of time for both pools to grow.
What’s interesting, though, is that both hinge on official World Health Organization (WHO) disease designations. To earn on a “Yes” bet on Polymarket, the WHO would need to declare a hantavirus-related outbreak a “pandemic,” the term the organization uses when a new illness, or a new strain of an illness, spreads worldwide. On Kalshi, meanwhile, the WHO would need to declare a “Public Health Emergency of International Concern,” another designation the organization uses to denote a significant health crisis with the potential to affect multiple countries.
This is the latest, and stark, reminder that betting markets turn longstanding institutions into arbiters of financial losses and gains. This is an obvious result of platforms that seek to financialize everything, but also seems to usher people back toward relying on central and trusted institutions, even amid an era when many of those institutions are facing attacks and record distrust.
This dynamic is true not just of the WHO—which the U.S. withdrew from earlier this year—but also of election officials, diplomatic negotiation teams, and even legislators who, by the nature of betting markets, have now become inadvertent referees of people’s financial fates – even if that has nothing to do with their actual purpose or operations. It’s also created a constellation of new perverse incentives that have pushed some employers, including the Senate, New York State’s government, and even JPMorgan to either caution or outright ban their staff from these platforms.
Notably, complaints obtained by Fast Company through a public records request show that sometimes, losers even bring their complaints about services like Kalshi to the Federal Trade Commission (FTC). At least a few of these complainers have seemingly asked the agency to intervene over the app’s adjudications of betting outcomes.
A spokesperson for Kalshi tells Fast Company that the markets are designed to help people and businesses quantify the risk of emerging public health threats. “They provide accuracy and clarity to people concerned about hantavirus that enables them to make better, more-informed decisions about how to proceed with their lives in the face of the significant risks it poses,” the spokesperson says. (Polymarket did not respond to a request for comment, and the FTC declined to comment.)
One person, for instance, reached out back in November 2024 to complain about a bet over whether Robert F. Kennedy Jr. would join the Trump administration, alleging that the interpretation of the bet changed after it had already been placed. As a result, the person claimed, they lost $2,000. After reaching out to the company, the person then contacted the FTC. Another person, also in 2024, complained to the Federal Communications Commission (FCC) about clarification issues surrounding bets on whether Elon Musk would join the government, and whether DOGE counted as a government organization.
Composite 56Trend 10
Canvas cyberattack disrupts final exams for colleges nationwide. Here’s what to know
Schools and universities across the country are recovering from an outage that knocked down Canvas, an online platform that manages exams, course notes, lecture videos, and grades. The disruption tied to a cyberattack hit in the middle of finals period for many colleges, a high-stress time when students and instructors rely heavily on the platform.
By late Thursday, Instructure, the parent company of Canvas, said the platform was available again to most users.
The hacking group ShinyHunters claimed responsibility for the breach, said Luke Connolly, a threat analyst at the cybersecurity firm Emsisoft. On Friday, Instructure and Canvas no longer appeared on a site where ShinyHunters lists its targets.
Some schools, however, have continued to block students and teachers from accessing Canvas, citing an abundance of caution while assessing security threats.
Here’s what to know about the outage.
What is Canvas?
Schools and universities use Canvas to manage nearly all aspects of instruction. The platform acts as a gradebook, a hub for digital lectures and course materials, a discussion board for classroom projects, and a messaging platform between students and instructors.
Some courses also give quizzes and exams on the platform, or use it as a portal where final projects and papers are submitted on deadline.
Who is ShinyHunters?
ShinyHunters is a loose association of teenage and young adult hackers in the U.S. and the United Kingdom who have been linked to other large-scale cyberattacks, including one on Ticketmaster, Connolly said. On the page listing their targets, the group describes itself as “rooting your systems since ‘19,” using a term for accessing a computer system’s deepest layer.
Earlier this week, ShinyHunters said that nearly 9,000 schools and 275 million individuals’ data could be leaked if schools did not pay the ransom by a deadline of May 6. The group then extended the deadline, indicating some schools had engaged with them to negotiate.
Schools and universities, rich in personally-identifiable information on students, teachers and employees, have become prime targets for criminal hackers in ransomware attacks. Targets can be individual districts, like the Minneapolis Public Schools or Los Angeles Unified School District, or external vendor platforms like Canvas or PowerSchool that education systems increasingly rely on to manage schedules, courses and exams.
The impact on students
Though most schools seem to have restored access to Canvas, the disruptions to finals period are likely to ripple throughout the week.
The University of Massachusetts at Dartmouth said that it would postpone exams scheduled for Friday and Saturday to ensure students had time to review course materials that would not have been accessible during the shutdown.
The University of Illinois postponed all exams that were scheduled to take place Friday, Saturday or Sunday for all classes, regardless of whether the courses utilized Canvas.
And Montgomery County Public Schools in Maryland continued to limit access to Canvas on Friday, citing an abundance of caution “while we work to better understand the full impact of the incident and any potential vulnerabilities involving information connected to the platform.”
___
The Associated Press’ education coverage receives financial support from multiple private foundations. AP is solely responsible for all content. Find AP’s standards for working with philanthropies, a list of supporters and funded coverage areas at AP.org.
—Annie Ma and Heather Hollingsworth, Associated Press
Composite 54Trend 10
OpenAI and Anthropic just met with religious leaders at the ‘Faith-AI Covenant.’ Here’s why
As concerns mount over artificial intelligence and its rapid integration into society, tech companies are increasingly turning to faith leaders for guidance on how to shape the technology — a surprising about-face on Silicon Valley’s longstanding skepticism of organized religion.
Leaders from various religious groups met last week with representatives from companies including Anthropic and OpenAI for the inaugural “Faith-AI Covenant” roundtable in New York to discuss how best to infuse morality and ethics into the fast-developing technology. It was organized by the Geneva-based Interfaith Alliance for Safer Communities, which seeks to take on issues such as extremism, radicalization and human trafficking. The roundtable is expected to be the first of several around the globe, including in Beijing, Nairobi and Abu Dhabi.
Tech executives need to recognize their power — and their responsibility — to make the right decisions, said Baroness Joanna Shields, a key partner in the initiative. She worked as a tech executive with stints at Google and Facebook before pivoting to British politics.
“Regulation can’t keep up with this,” she said. But the leaders of the world’s religions, with billions of followers globally, have the “expertise of shepherding people’s moral safety,” she reasoned. Faith leaders ought to have a voice, Shields said.
“This dialogue, this direct connection is so important because the people who are building this understand the power and capabilities of what they’re building and they want to do it right — most of them,” she said of AI tech executives.
The goal of this initiative, according to Shields, is an eventual “set of norms or principles” informed by different groups and faiths, from Christians to Sikhs to Buddhists, that companies will abide by.
Challenges lie ahead
Present at the meeting were a variety of faith groups, including representatives from the Hindu Temple Society of North America, the Baha’i International Community, The Sikh Coalition, the Greek Orthodox Archdiocese of America and The Church of Jesus Christ of Latter-day Saints, widely known as the Mormon church.
Before these companies initiated outreach, some traditions had issued their own ethical guidance on using AI. The Church of Jesus Christ of Latter-day Saints has given a qualified approval of the technology in its handbook. “AI cannot replace the gift of divine inspiration or the individual work required to receive it. However, AI can be a useful tool to enhance learning and teaching,” it reads.
The Southern Baptist Convention, the largest Protestant denomination in the U.S., passed a resolution in 2023: “We must proactively engage and shape these emerging technologies rather than simply respond to the challenges of AI and other emerging technologies after they have already affected our churches and communities.”
One challenge in creating a list of common principles is that global faiths, despite common ground, differ in their values and needs. “Religious communities see priorities differently,” said Rabbi Diana Gerson, a roundtable participant and the associate executive vice president of the New York Board of Rabbis.
The partnership highlights a growing coalition between faith and tech, born out of an effort to create moral AI — a contested concept which begs questions about whether that is possible and what it means.
“We want Claude to do what a deeply and skillfully ethical person would do in Claude’s position,” Anthropic states in the public “Claude Constitution” written for its chatbot. That constitution was made with the help of a host of religious and ethics leaders.
In this burgeoning alliance, Anthropic has been the most assertive, at least publicly, in their efforts to court faith leaders. The move follows a public dispute earlier this year with the Pentagon over military use of artificial intelligence after Anthropic said it would restrict its technology from being used to develop autonomous weapons or for mass surveillance of Americans.
“There’s some aspect of PR to it. The slogan was ‘Move fast and break things.’ And they broke too many things and too many people,” said Brian Boyd, the U.S. faith liaison for the nonprofit Future of Life Institute. “There’s both a moral obligation on the part of the companies that they’re belatedly recognizing, as well as I think, for some members of the companies, an earnest questioning.”
Some skepticism emerges
But other advocates for AI regulation and safety aren’t so sure these efforts are genuine.
“At best it’s a distraction. At worst it’s diverting attention from things that really matter,” said Rumman Chowdhury, the CEO of the nonprofit Humane Intelligence and the U.S. science envoy for AI under the Biden administration.
Chowdhury says she’s not inclined to believe religion is the best place to help answer questions surrounding AI and ethics, but thinks she understands why companies are increasingly turning to it.
“I think a very naive take that Silicon Valley has had for a couple of years related to generative AI was that we could arrive at some sort of universal principles of ethics,” she said. “They have very quickly realized that that’s just not true. That’s not real. So now they’re looking at maybe religion as a way of dealing with the ambiguity of ethically gray situations.”
It’s unclear to what extent these notoriously opaque companies are translating what they hear from faith leaders into action — and what that action might look like. But some critics fear the conversation about creating ethical versions of the technology distract from broader conversations about AI and its role in society.
“Under the guise of, ‘We’re gonna build all this stuff. That’s a given. And when we do build these things in these ways, how do we make sure that the end result is maybe good,'” said Dylan Baker, the lead research engineer at the Distributed AI Research Institute. “It’s like, ‘Wait, wait, wait. We need to question whether we want to be building these things at all.”
Associated Press religion coverage receives support through the AP’s collaboration with The Conversation US, with funding from Lilly Endowment Inc. The AP is solely responsible for this content.
—Krysta Fauria, Associated Press
Composite 56Trend 10
This AI startup wants to help smooth complex industrial materials sales
While many AI companies are betting their products can be useful to a broad segment of businesses, a startup called Emanate is taking the opposite approach, building highly targeted tools designed for complex sales transactions in the industrial materials sector.
Founder and CEO Kiara Nirghin says the somewhat esoteric market, which includes manufacturers, distributors, and service providers working with materials from steel building materials to metal piping, has intricate sales processes involving generating quotes for bespoke orders, connecting existing customers with goods they may need, and proactively finding new customers.
The industrial materials sector, which provides raw materials like steel and aluminum and manufactured parts like wire and pipe, is vital to both the push to boost U.S. manufacturing output and the shift to a greener economy, which itself requires manufacturing solar panels, wind turbines, and electric vehicle charging stations. The metals and minerals industry alone is a multi-trillion dollar sector, and Emanate argues quickly generating more precisely quotes and closing sales faster can boost productivity and reduce waste from mistargeted production.
But right now, even generating quotes with existing systems can take as long as three to four weeks, says Nirghin, and until recently AI
systems weren’t sophisticated enough to take over for humans. Now, she says, they can generate useful quotes close to instantaneously.
“That was only recent—in terms of the last approximately six to eight months,” she says. “So there is a very big change in quality and step function in terms of actually applying the models.”
But, says Nirghin, the real key isn’t the underlying AI models but the so-called harness—the framework of AI-callable tools, integrations with other systems like enterprise resource planning (ERP) software and databanks of corporate knowledge, and custom configurations—that wrap around them to form AI agents.
Emanate, which has received funding from investors including Andreessen Horowitz and M13 (though Nirghin declined to disclose the exact amount of funding the San Francisco-based company has raised) and currently has 10 employees, is explicitly betting that markets like industrial materials will benefit from sector-specific AI tools rather than simply adopting standard, off-the-shelf AI agents.
Setting up Emanate’s system for a new customer isn’t simply a matter of activating a chatbot. It’s a process that can take from eight to 12 weeks, including identifying critical data sources from ERP databases to repositories of past sales email correspondence and PDFs containing valuable data, then getting set up to securely connect to them. Once the system is set up, customers can also continue to build upon and customize the AI agents involved, says Nirghin.
The specialized approach is designed for greater accuracy than general-purpose AI tools, and the company also works with its customers to track data points like number of quotes processed, hours spent by human workers, sales leads handled, and outbound messages sent before and after the technology is deployed. And while some other AI companies more heavily focus on helping customers cut costs through automation, Nirghin says Emanate is focused on revenue growth, aiming to give its customers a revenue boost of 40% or more.
“We capture a full baseline before we go live, and then we track every metric,” Nirghin says. “We’ve been very clinical in measuring these metrics so that we can actually report and communicate on them.”
Naturally, at the start of a new deployment, humans typically also review quotes and messages generated by the AI before they’re sent to customers. But over time, they’re typically willing to defer more to the AI.
Nirghin, who has previously received support from a fellowship program run by Alexis Ohanian’s 776 Foundation and the Thiel Fellows program, says she believes the company’s specialization and industry focus will give it a sustainable advantage in catering to the industrial materials sector as it works to meet the needs of a growing U.S. manufacturing economy. The AI’s success at helping secure deals can even help boost production and employment among materials companies, many of which have the capacity to grow as they secure customers, she says.
But in the future, she says, the same approach could serve other industries with similar specialized sales and distribution needs, she says, including the electrical and chemical industries.
“That is obviously our broader vision, and what our investors obviously get really excited about as well,” she says.
Composite 53Trend 10
How Ask Jeeves blew it
Hello again, and welcome back to Fast Company’s Plugged In.
Upon hearing of a celebrity’s death, have you ever been startled to realize that they hadn’t left us long ago? That happened to me last weekend. Except the dearly departed in question wasn’t a person, but a company: Ask.com, the web property forever better known by its original brand, Ask Jeeves.
For years, I wrote about Ask quite regularly. But when its owner, media conglomerate IAC (which is in the process of changing its own name to People Inc.), announced it had shut down the site as of May 1, it was its first time in the news in more than 15 years. The last time before that was in November 2010, when IAC gave up on Ask being a general-purpose search engine and turned it into a user-generated Q&A site. At some point in between those two moments, Ask had morphed into a bottom-feeding portal for articles so out of date that “10 Best Documentaries of 2022—So Far” was one of the headlines on its homepage when IAC pulled the plug.
In other words, it’s been a long time since Ask.com mattered. And yet its demise inspired a flurry of nostalgic reveries, focused on its early days, original name, and cartoon butler mascot. That residual fondness reminded me that once upon a time, the company really had something. But instead of capitalizing on what it had created, it gave up—just before it might have been able to fulfill its vision.
Ask Jeeves debuted in 1997, a moment of great expectations for the nascent field of internet search. As the web exploded with content, Ask Jeeves was one of a bevy of startups that emerged to organize it. Yahoo and AltaVista were the big dogs, but others included Excite, Lycos, HotBot, LookSmart, Northern Light, and WebCrawler.
A very early version of Ask Jeeves, when the entire internet looked like a junior high student’s GeoCities page.
Meanwhile, a couple of Stanford graduate students, Larry Page and Sergey Brin, were working on their own search algorithm. When Google launched in 2008, its results were clearly the best in the business, and its ascent was rapid. In 2001, Ask Jeeves responded by buying a startup called Teoma, whose relevance-ranking algorithm was a credible rival to Google’s PageRank. The move certainly felt like a sizable whoop at the time. Or at least it wasn’t yet a given that Google’s momentum was unstoppable.
In 2003, however, Google overtook Yahoo as the dominant search site. After that, there was never a moment when Ask Jeeves, or anyone else, was poised to catch up. Google’s market share steamrolled to 90%-plus, leaving its rivals squabbling over what little remained.
The Ask Jeeves homepage didn’t have an “I’m feeling lucky” button. But it did have a butler.
But even after IAC took control of Ask Jeeves—the conglomerate bought it for $2 billion in July 2005 and quickly eliminated the “Jeeves” from its name—you couldn’t accuse the site of doing too little in search of success. Instead, it was all over the place, flinging new ideas at the wall and barely waiting to see if they stuck before moving on to new ones. In June 2007, it released an all-new design that offered tons of useful features Google lacked at the time. By October of the following year, however, it had dumped many of them in favor of an experience that felt like warmed-over Google.
As an IAC property, Ask advertised constantly on TV, but never landed on a brand promise that stuck. At one point, its commercials positioned the site as being for serious searchers who craved advanced tools. Then they claimed it offered “instant getification.” Sometimes they didn’t offer any reason to try it beyond the fact that it wasn’t Google.
All along, I rooted for Ask, simply because even hapless competition for Google served consumers better than no competition at all. But it floundered so publicly that it wasn’t surprising when IAC downsized it to a mundane Q&A platform almost 16 years ago.
Okay, back to 2026 and the eulogies inspired by Ask.com’s shuttering. As far as I can tell, nobody ever cherished that brand. But boy, did Ask Jeeves and its butler lodge themselves in people’s brains. The vast majority of headlines mentioned both, more than 20 years after they putatively entered retirement. (IAC did bring back Jeeves in the U.K. in 2009, in a more dynamic computer-rendered version who bore an eerie resemblance to its chairman, Barry Diller—or at least I thought so at the time.)
Jeeves as he appeared in 2009, when he returned to service in the U.K. with a new suit and slightly smug expression.
In its pre-IAC period, Ask Jeeves bet big on the appeal of its affable, balding mascot, who it maintained was unrelated to writer P. G. Wodehouse’s legendarily capable manservant, though it added a credit to its homepage after the Wodehouse estate complained. A company representative told Salon’s David McDonough that it wanted to make the character as familiar as Popeye. In 1999, Jeeves rode on a float in the Macy’s Thanksgiving Day Parade; the following year, he was upgraded to full balloon status.
If you’d compiled a list of the internet’s most familiar fictional characters around the turn of the century, Jeeves would have been on it, along with the dancing baby, the Pets.com sock puppet, and BonzaiBuddy. Apparently IAC preferred a more modern, less whimsical image for its search engine. Still, when it did away with Jeeves, it torched a massive amount of brand equity.
Ask also failed to build on its original potential in a more fundamental way. Ask Jeeves’s very name suggested that it wasn’t about searching the World Wide Web so much as getting answers to questions. Back then, it was a fuzzy distinction, since the answers you sought were generally scattered across the web. But even as IAC was exiting the search business, Google was working on a technology called the knowledge graph. When it appeared, in 2012, it dramatically increased the percentage of questions the search engine could answer without routing users to other sites. Ask Jeeves could have offered similar features had it remained in the game.
If the site had held on as a search engine all the way into the generative AI age, it might have become the product it always aspired to be: an engaging, hyper-knowledgeable assistant with an uncanny ability to field questions on any topic. Today, Jeeves could also help us manage our calendars, buy stuff, and take care of personal and professional business far outside the realm of 1990s search engines. He could be the ultimate AI agent—and being personified as a cartoon butler would make perfect sense. (In 2023, Ask Jeeves cofounder Garrett Gruener told The Atlantic’s Charlie Warzel that he was proud of the product’s prescience and didn’t feel too bad about losing the search wars to Google.)
As I was mulling over what might have been, it dawned on me that even if IAC failed to seize the opportunity to infuse Jeeves with AI, I could. Chatbots are adept at role-playing, a fact that is often disturbing. But their willingness to take on a persona let me whip up a prompt to turn any bot into a butler.
Voilà:
“Until I request otherwise, take on the role of Jeeves, an experienced, helpful, extraordinarily competent British butler. Respond to my prompts in a dignified, slightly reserved manner that is deferential but not obsequious. Behave as if you are a salaried employee but also sincerely concerned about looking out for me. Use information you know about my interests and habits to facilitate efficient and thoughtful responses. Decline to undertake any requests that are inappropriate.”
Plugging in these instructions to ChatGPT, Claude, Gemini, and Copilot got me entertaining results—especially in the case of Claude, whose stock personality is crisp and professional in the first place. I don’t plan to use them forever, but resuscitating Jeeves for a few days seems like an appropriate way to mourn one of the 20th-century internet’s true giants. If you’re similarly inclined, give them a try in your favorite chatbot, and let me know what you think.
You’ve been reading Plugged In, Fast Company’s weekly tech newsletter from me, global technology editor Harry McCracken. If a friend or colleague forwarded this edition to you—or if you’re reading it on fastcompany.com—you can check out previous issues and sign up to get it yourself every Friday morning. I love hearing from you: Ping me at [email protected] with your feedback and ideas for future newsletters. I’m also on Bluesky, Mastodon, and Threads, and you can follow Plugged In on Flipboard.
More top tech stories from Fast Company
A PC trade-in rush is on the way—and it’s coming at the worst possible time
As millions of pandemic-era PCs near the end of their lifespan, consumers are running into soaring hardware prices driven by the AI boom. Read More →
Grok’s usage is so low that Elon Musk can sell compute to Anthropic
Anthropic says it’ll use all the AI compute capacity from SpaceX’s ‘Colossus 1’ data facility in Memphis. Read More →
Bose is rebooting its smart speakers for the Sonos haters
The audio giant spent years reviving its classic Lifestyle speaker line, with hopes of making it future-proof. Read More →
OpenAI’s trillion-dollar AI bet is a study in ‘riskmaxxing’
The AI giant is betting its future on a rapid increase in demand for frontier AI models in the coming years. Read More →
Chinese humanoids are leaving American robots in the dust
Asia is spending billions on the development and deployment of humanoids that are already taking on humans’ least desired jobs. Read More →
AI? No thank you! 3 truly free, no-AI apps for the overwhelmed
Pick up a tool that does exactly one thing and then gets out of your way—no LLM involved. Read More →
Composite 56Trend 10
Kalshi’s $22 billion problem
Grant Mainland had a tough day at the office earlier this week. A lawyer representing the prediction markets platform Kalshi, Mainland appeared before the Massachusetts Supreme Judicial Court on May 4 with an unenviable task: persuading the justices that a company that has literally advertised itself as the “first app for legal sports betting in all 50 states” is not, technically speaking, offering people the opportunity to bet on sports.
Mainland was hoping to get the court to overturn a lower state court injunction that blocked Kalshi from offering its “markets” related to sports within the commonwealth’s borders. Thanks primarily to these sports markets, which accounted for nearly 90% of its revenue in 2025, Kalshi has hit $1.5 billion in annualized revenue.
It’s a growth story that investors are clearly buying: Kalshi recently announced it raised a $1 billion Series F, catapulting the company to a $22 billion valuation—double what it was worth just six months ago.
For the uninitiated, Kalshi allows users to make money by correctly predicting the yes-or-no outcomes of real-world events. Users are able to buy and sell contracts at prices that range from 1 cent to 99 cents, which roughly approximate the market’s sense of the percentage chance that an outcome will occur. When that market “resolves” (i.e., the event either happens or doesn’t), those who hold shares in the winning position are paid out at $1 per share. If, for example, my beloved Seattle Seahawks return to the Super Bowl next year, anyone who bought in at 14 cents per share—the price as of this writing—will enjoy a nice payday.
If this sounds to you like futures betting by another name, you’re not alone. In January, a lower court judge found that Kalshi, by allowing users to buy and sell “event contracts” on everything from final scores to player props, was functionally operating in Massachusetts as an unlicensed sportsbook. There is “no question,” Judge Christopher Barry-Smith wrote, that requiring Kalshi to follow the same laws as every other sportsbook—and putting it on ice in the meantime—would serve “both public health and safety, and the Commonwealth’s financial interest.”
Mainland’s primary arguments in Commonwealth of Massachusetts v. KalshiEX LLC are the same arguments Kalshi always makes when pressed about the sports side of its business: that as an exchange regulated by the federal Commodity Futures Trading Commission, Kalshi is not subject to state regulation. It also contends that its products are not “bets,” but “swaps,” a type of derivative contract that companies have long used to hedge against financial risk.
Things did not go smoothly for Mainland when he made this case before Massachusetts’s highest court. He was quickly interrupted by Justice Gabrielle Wolohojian: “If we just zoomed up one level, ‘event contracts’ would not be conceptually incompatible with what we would historically understand to be a bet or wager.”
When Mainland asserted that buying and selling contracts on Kalshi is “completely different” from laying a wager with a sportsbook, Justice Scott Kafker sounded baffled: “Completely different? For someone who wants to bet on a game, this is a way of betting on a game, right?”
Mainland tried to forge ahead, but Kafker could not conceal his skepticism. “I understand you can distinguish it,” he said to Mainland. “But if I want to bet on this stuff, I can do it this way, too.” At one point, Kafker characterized Mainland as “swimming upstream here,” which, as a lawyer, is not what you want to hear a judge say about your legal argument.
The oral argument in Commonwealth v. KalshiEX is part of a national trend in which states, at last aware that prediction markets are depriving them of tax revenue and opening up de facto sports betting to people who might still be in high school, are trying to reassert themselves a bit. These legal fights pit states against billion-dollar companies and a Trump administration with a vested interest in ensuring prediction markets’ continued profitability.
Massachusetts is one of many states that have sued Kalshi in recent months for alleged violations of state gambling laws. On April 3, Nevada regulators celebrated when a state court judge issued an injunction banning the company from offering sports contracts, which he described as “indistinguishable” from placing a bet, in the state.
Others have been even more aggressive, creative, or both in their enforcement efforts. Arizona’s attorney general filed criminal charges against Kalshi, accusing it of running an illegal gambling business. Lawmakers in Utah passed a law to ban prop bets, which is a little odd, given that the state already prohibits gambling (in fact, it’s part of the state constitution). In a February op-ed in Deseret, though, Utah’s attorney general implied strongly that the ban on prop bets specifically targets prediction markets like Kalshi, and that he would use it to go after those companies as soon as the governor signs off.
The principal challenge these attacks face is that Kalshi, which during football season does 90% of its volume on sports contracts, has invested lots of time and money preparing to fend them off. In January 2025, shortly after Donald Trump’s inauguration, the company announced that the president’s eldest son, Don Jr., had joined the company as a strategic adviser. A few months later, he took a similar position at Polymarket.
Earlier this year, Kalshi blanketed downtown Washington, D.C., in splashy mint-green ads assuring commuters that the platform is safe and legitimate. The tone of the campaign is unmistakably urgent, in the “doth protest too much” sense of the word; as Fast Company’s Joe Berkowitz pointed out, if you are a business that still feels compelled to make crystal clear that you “operate under U.S. law,” that’s a sign that the PR department has a lot of work to do.
[Photo: Daniel Heuer/Bloomberg/Getty Images]
Fortunately for Kalshi, the Trump administration has not required much persuasion. Although the president has occasionally criticized prediction markets, his media company is working on launching its own prediction market for users of his social media platform, Truth Social.
In a wild coincidence, CFTC Chair Michael Selig, whom Trump nominated in October 2025, has aggressively defended his jurisdiction over prediction markets—so much so that critics have described him as less a “normal regulator” than a “cheerleader for the industry.” During the Biden administration, the CFTC proposed a rule that would have banned event contracts related to politics and sports; shortly after taking office, Selig withdrew it.
At least some judges have come down on Selig’s and Kalshi’s side: In early April, a federal appeals court found that the federal Commodity Exchange Act indeed preempts state gambling laws, allowing Kalshi to operate in New Jersey over the objections of state officials. On X, Selig applauded the court for its “decision to uphold federal law.” This week, a federal district court judge ended Arizona’s criminal prosecution of Kalshi, which he said would create an “inconsistent regulatory patchwork that Congress intended to avoid.”
Other courts, however, have remained leery: In April, a three-judge panel of the 9th Circuit Court of Appeals sounded reluctant to intervene on Kalshi’s behalf in its dispute with Nevada regulators; one judge, Bridget Shelton Bade, remarked that based on Kalshi ads that she encounters “almost every day” on her phone, “it seems like they are advertising this as sports betting.”
During oral argument in Maryland’s litigation against Kalshi this week, a 4th Circuit judge invoked the classic farm animal analogy: “If it quacks, you know, it’s a duck, right?” Judge Roger Gregory said.
What state gaming regulators are really after here, of course, is tax revenue; of the billions of dollars in sports-related volume that Kalshi does each year, states do not collect any of it. But there is also a growing body of evidence that sports event contracts inflict real-world harms on the users whom regulators are supposed to protect—harms that anyone familiar with this country’s sports betting boom will recognize.
In just about every meaningful way, Kalshi operates like a conventional sportsbook: It is available on smartphones, for example, and nudges winners riding a dopamine high to play again. Yet Kalshi is not subject to state laws that prohibit sportsbooks from taking bets from people under 21, and that require sportsbooks to take specific steps to discourage problem gambling and prevent insider betting.
In Massachusetts (like in most states) sportsbooks like DraftKings and FanDuel must participate in a system that allows users to voluntarily exclude themselves from licensed betting platforms in the state. But Kalshi is not licensed, which means that a Massachusetts resident who is struggling with compulsive gambling, and who opts into the system in an effort to stop, will be as free as ever to open up the Kalshi app and place another bet.
A recent Wall Street Journal analysis sheds some light on just a few consequences of a status quo in which functionally identical prediction markets can operate in parallel with state-licensed sportsbooks. Although prediction markets pitch themselves as a way to make easy money, in reality a tiny fraction of sophisticated professionals take home most of the winnings.
Meanwhile, ordinary people are losing eye-popping amounts of money on, for example, whether A$AP Rocky says the word rapper during a Tonight Show interview with Jimmy Fallon. The evaporation of your life savings is not any less devastating if you lose it on a contract purchased on a federally regulated exchange, instead of a bet placed at a state-regulated sportsbook.
The basic question that lawmakers and courts are grappling with right now is less legal than it is philosophical: whether to classify Kalshi’s business based on how the company organizes itself, or on how it appears to consumers and works in the real world. To date, Kalshi has been mostly able to maintain its position of privilege in the regulatory landscape.
But as oral argument in the Massachusetts case suggests, for an increasing number of people in positions of power, the distinction between money-line wagers and event contracts is no longer meaningful, to the extent that it ever was in the first place.
Composite 56Trend 10
The AI productivity playbook: tips and tools for you
Fast Company’s global tech editor Harry McCracken and tech writer Jared Newman cut through the AI hype to walk you through the tools and techniques that are making a difference in the way they work. In this conversation, they break down the trends behind 2026’s most forward-thinking organizations and share the practical, steal‑worthy strategies that leaders at all levels can apply right now. Whether you’re refining your road map or scanning the horizon for what’s next, their overview will provide you with actionable insights and valuable new perspectives.
Composite 58Trend 10
Google used to be a search engine. Now it wants to be everything
Twenty years ago, if you asked the average person what Google was, they’d tell you it was a search engine. The company became synonymous with searching for information online, reaching a level of dominance no search engine had seen before, or has seen since.
Ask the average person today and they’d probably tell you the same thing. Except Google isn’t just that anymore. It’s a far more complicated company, one trying to be all things to all people, and arguably succeeding at none of them.
Google is now a five-layer company, says David Bader, director of the Institute for Data Science at the New Jersey Institute of Technology. One of the key layers is AI, which could account for $185 billion in capital expenditure this year, “larger than the GDP of most countries,” according to Bader. That level of spending signals how dramatically the company has changed direction. “No serious search-only company spends like this,” he says.
That focus on AI is increasingly visible to end users, with AI layered into more and more Google products. “They’re shoving Gemini into every nook and cranny, whether it’s GSuite, whether it’s email, whether it’s Maps, whatever,” says Alex Hanna, a former Google employee and director of research at the Distributed AI Research Institute.
Still, there remains a gap between what Google is, depending on who you ask and where they sit. “There’s how Google sees itself internally, which I think is they see themselves a bit more as an AI company,” says Hanna. That contrasts with how much of the world still sees the company: primarily as a search engine. And, in Hanna’s view, that experience has deteriorated. “When you use Google Search, it’s trash. It sucks,” she says.
Hanna argues that the decline in search quality is partly tied to the way Google is reshaping its business model for the post-ChatGPT era, one in which AI can bypass traditional search entirely and reduce the need for users to visit either a search engine or the websites it indexes.
Advertising remains Google’s “cash cow,” says Bader, accounting for 74% of its revenue. But others believe that dominance could erode as AI reshapes search behavior. “They know that what they have to move to is a model that isn’t based on ad revenue,” says Hanna. “It’s based on whether they can find a pathway to monetize the AI infrastructure that they’ve been building out.”
Still, “Google Search isn’t going away,” says Gartner analyst Ed Anderson. “And I think Google Search will continue to be one of the primary touchpoints for years to come.”
Beyond monetizing its AI infrastructure, Google is also reshaping other parts of its business to maintain its cash flow. That includes generating billions through cloud infrastructure, which Bader says has grown 63% year over year and has become “a real number three to AWS and Azure,” accounting for roughly a fifth of the company’s overall business.
Google is also increasingly deploying its capital as an investor. The company owns about 6% of SpaceX and roughly 14% of Anthropic, alongside stakes in or ownership of companies including Waymo and Wiz, the latter of which it acquired for $32 billion, plus dozens of other holdings. The result, some critics argue, is that Google increasingly resembles “a glorified venture capital fund.”
Whether having so many fingers in so many pies means Google has lost its way, or simply found a new one, remains an open question. “We live in an economy where you have to show growth,” says Hanna. “They’re in this very weird position where they are in third place or fourth place again, even though they were a first mover on the tech.” That, in turn, makes the company feel even more muddled.
Not everyone is so pessimistic. “The interesting part is not that any single label [of what Google is] is wrong,” says Bader. “The interesting part is that all five are simultaneously true, and that’s never been true of any single company before.”
Composite 58Trend 10
If you’re looking for a modern BlackBerry-style phone, this is the one to beat
BlackBerry revivalist phones have been appearing in various forms over the last few years, but the Unihertz Titan 2 Elite is the most credible option yet. The small-scale Chinese boutique-of-sorts Unihertz has spent years refining its formula to balance modern Android capabilities with legacy tactile hardware. In 2026, it’s finally landed on a device that makes the most of its own identity.
The naming convention here is admittedly a little confusing. Last year’s Titan 2 was a rugged, wide-format device clearly inspired by the BlackBerry Passport—it was, in every sense, “titanic.” But this new Elite successor isn’t a turbo-charged version of that phone; it’s a completely different animal. It ditches the ultra-wide, ruggedized footprint for a much smaller design that feels like a spiritual successor to the BlackBerry Q10.
A departure
The Elite 2’s industrial design is a departure from Unihertz’s recent “brick” aesthetic. This isn’t a rugged phone, and it feels better for it. That said, I would personally avoid the iPhone 17 Pro-inspired orange model, which seems unlikely to age too well; the black model is much more in keeping with the BlackBerry heritage.
{"blockType":"mv-promo-block","data":{"imageDesktopUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/12\/multicore.png","imageMobileUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/12\/multicore-mobile.png","eyebrow":"","headline":"\u003Cstrong\u003E\u003Cstrong\u003ESubscribe to Multicore\u003C\/strong\u003E\u003C\/strong\u003E","dek":"Multicore is about technology hardware and design. It’s written from Tokyo by Sam Byford. To learn more, visit \u003Ca href=\u0022https:\/\/www.multicore.blog\/\u0022\u003Emulticore.blog\u003C\/a\u003E","subhed":"","description":"","ctaText":"SIGN UP","ctaUrl":"https:\/\/www.multicore.blog\/","theme":{"bg":"#f5f5f5","text":"#000000","eyebrow":"#9aa2aa","subhed":"#ffffff","buttonBg":"#000000","buttonHoverBg":"#3b3f46","buttonText":"#ffffff"},"imageDesktopId":91454027,"imageMobileId":91454030,"shareable":false,"slug":"","wpCssClasses":""}}
At 10.6 millimeters thick, the Elite 2 is substantial by modern standards, but that’s still only about as thick as a classic BlackBerry; the compact footprint (just 117.8 millimeters tall) also keeps it from feeling overbearing in a pocket. Unlike the Titan 2, there’s virtually no bezel above the display.
The superfluous secondary screen found on the back of the original Titan 2 has also been removed, helping to achieve a cleaner, more focused look. And the front of the phone is split between the screen and the keyboard, with virtually no wasted space anywhere else.
The display is a 4.03-inch 120-hertz OLED panel, which gives a significant step up in saturation and contrast from the LCDs that Unihertz has used in the past, though the squarish aspect ratio means video content and social media scrolling aren’t quite its strong suit. The panel also exhibits significant color shifting when viewed at an off-angle, betraying its budget nature.
The keyboard, however, is excellent. It’s smaller than the one on the Titan 2, but the tactile response feels far more consistent and satisfying. It uses a four-row layout with more even backlighting and capacitive touch support, allowing you to scroll through content simply by swiping your thumb up and down the physical keys. The software integration is quite deep; you can program a long-press on “T” to open TikTok or “X” for X, for example, and there’s a dedicated “Action Button” on the side of the phone that allows for further customization.
One of the smartest changes is the layout. Unihertz moved the standard Android navigation keys to the bottom row alongside the space bar. This setup is more like the 2013 BlackBerry Q10 than the classic earlier devices, and it feels much more intuitive for a modern Android phone that doesn’t need the dedicated physical call/hang-up buttons of the 2010s.
I’ve been using the Standard model of the Titan Elite 2, which features a MediaTek Dimensity 7400 chip; there will also be a Pro version later in the year with a Dimensity 8400 chip. Personally, I’ve found the 7400 to be more than adequate for the kind of tasks I’d want to use a phone like this for. Both models also come with 12 gigabytes of RAM, which is plenty for the typical use cases.
The 4,050-milliampere-hour (mAh) battery has reasonable endurance, but I did notice fairly aggressive standby drain when the phone wasn’t in use. There’s also no wireless charging, though you can power the phone over a cable at up to 33 watts.
The software
On the software side, the Titan Elite 2 runs a very clean version of Android 16. Unihertz claims it will provide five years of updates, which is a strong commitment for a niche brand but perhaps not something you should treat as a surefire promise.
This is still a $400 phone from a smaller manufacturer, and it has the quirks to prove it. There is a generic “NFC” logo emblazoned on the camera bump that feels entirely unnecessary, and the camera system itself won’t be winning any awards for tasteful processing or low-light performance. The software is also extremely bare-bones, which will appeal to some Android purists, but falls some way short of the sleek, native software found on most true BlackBerry handsets.
But I do think Unihertz has finally nailed the form factor with the Titan Elite 2. It’s a fun, intentional device that doesn’t try to be everything to everyone, and I think it will be more appealing to most people who are interested in this sort of thing in the first place. It’s accessible where its predecessor was aggressive.
At this point, Unihertz has more experience building physical keyboard phones than almost anyone else left in the industry, and it shows; the Titan Elite 2 is by far its best phone to date. It comes at a time when the market is getting a little more crowded with upstart competition like the Clicks Communicator. For now, though, if you’re looking for a modern BlackBerry-style phone, this is the one to beat.
{"blockType":"mv-promo-block","data":{"imageDesktopUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/12\/multicore.png","imageMobileUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/12\/multicore-mobile.png","eyebrow":"","headline":"\u003Cstrong\u003E\u003Cstrong\u003ESubscribe to Multicore\u003C\/strong\u003E\u003C\/strong\u003E","dek":"Multicore is about technology hardware and design. It’s written from Tokyo by Sam Byford. To learn more, visit \u003Ca href=\u0022https:\/\/www.multicore.blog\/\u0022\u003Emulticore.blog\u003C\/a\u003E","subhed":"","description":"","ctaText":"SIGN UP","ctaUrl":"https:\/\/www.multicore.blog\/","theme":{"bg":"#f5f5f5","text":"#000000","eyebrow":"#9aa2aa","subhed":"#ffffff","buttonBg":"#000000","buttonHoverBg":"#3b3f46","buttonText":"#ffffff"},"imageDesktopId":91454027,"imageMobileId":91454030,"shareable":false,"slug":"","wpCssClasses":""}}
Composite 25Trend 10
Saudi Arabia dumped LIV—but it’s not getting out of the golf world entirely
In late April, indoor golf and entertainment brand Five Iron Golf launched its first location in Saudi Arabia after nearly three years of coordination with Golf Saudi, an affiliate of the country’s massive sovereign wealth fund.
The golf center, located on the ground floor of the Public Investment Fund Tower—the fund’s headquarters and the tallest building in Riyadh—offers a similar mix of simulators, leagues and lessons, and food and music as other Five Iron locations both in the U.S. and abroad, says Jared Solomon, cofounder and CEO of the brand.
“People are hitting golf balls, they’re having fun, they’re eating food,” he says. “They’re not drinking, because there’s no drinking there, but they’re having a great time.”
Five Iron Golf, which launched in Manhattan in 2017 (the name nods to both the club and Fifth Avenue in the Flatiron District, where it opened its first location), now has more than 40 U.S. venues. The company has expanded abroad through franchise deals in Singapore, Dubai, Spain, and Portugal.
The Saudi location opened under a similar arrangement with Golf Saudi. Solomon declined to share details, but says franchise agreements typically include fees of $50,000 to $100,000 per location and about 6% of revenue.
Solomon hopes the PIF Tower site will be the first of multiple locations in Saudi Arabia. But the Five Iron launch wasn’t the biggest golf news out of the region this spring. That distinction belongs to the announcement by PIF that it would no longer fund the controversial LIV Golf tour, which has received more than $5 billion in PIF backing since its 2022 launch.
In a widely circulated statement, PIF said backing LIV Golf no longer aligns with its current investment strategy, pointing to “investment priorities and current macro dynamics.” (The fund didn’t respond to Fast Company’s requests for comment.)
“The bone saw tour”
The move follows the Saudi-backed LIV tour’s disruption of the golf world, drawing players away from the gold-standard PGA Tour, which banned players from competing in both. It also raises broader questions about the future direction of the trillion-dollar PIF, which has invested heavily in global tech companies (from Uber to Magic Leap) and through vehicles like the $100 billion SoftBank Vision Fund. It’s also set to acquire a substantial stake in a merged Paramount-Warner Bros. Discovery media business.
Under the leadership of Crown Prince Mohammed bin Salman, Saudi Arabia has made clear its ambitions under its Vision 2030 project—a sweeping state-led plan to remake the kingdom’s economy—to transition from an oil-based economy to a hub for business, sports, tourism, and entertainment, powered in part by PIF investments. The country will host the 2034 FIFA World Cup and a 2030 world’s fair-style expo, and PIF recently reaffirmed its commitment to Newcastle United Football Club, maintaining its majority ownership of the Premier League team.
Those investments are not purely economic. They’re also intended to reshape the kingdom’s global image after years of scrutiny over human rights violations, limits on women’s rights, and the 2018 murder of journalist Jamal Khashoggi inside the Saudi consulate in Istanbul.
“The objective of all of this sports funding is twofold,” says Aaron Ettinger, a political science professor at Carleton University in Ottawa. “One is to make money, and the other one is political. The political objective is to make Saudi Arabia look like a normal country and draw international investments and all that kind of stuff.”
The decision to pull back from LIV may reflect the tour’s failure to deliver on either front. As Ettinger notes, upstart leagues often struggle to challenge entrenched competitors like the PGA. (Indeed, recent U.S. sports history is full of failed competitors to the National Football League.) LIV relied heavily on Saudi funding, and it’s not clear that it did anything to boost the country’s image abroad.
“It was referred to as the ‘bone saw tour’ when it got started,” Ettinger says. “It just ended up drawing more attention to the kind of stuff about Saudi Arabia that makes people in the West really uncomfortable.”
The LIV pullback also comes as the U.S.-Israeli war with Iran has disrupted global oil markets, and as Saudi Arabia has stepped back from other ambitious, high-cost projects, including a planned linear city and a massive cubical building in Riyadh.
“I think even before the war started, there were already indications that the PIF was reconsidering how sustainable some of its investments are over the long term,” says Andrew Leber, an assistant professor in Tulane University’s political science department and a non-resident fellow at the Middle East Program within the Carnegie Endowment for International Peace.
In mid-April, PIF unveiled a new five-year strategy, marking a shift from “rapid growth and acceleration” to a phase centered on sustained value creation, with an emphasis on efficiency, impact, and stronger governance. Leber says that likely means a renewed focus on domestic investments tied to employment and political stability, and less emphasis on speculative global ventures.
The pivot also coincides with rising tensions between Saudi Arabia and the United Arab Emirates, which recently announced plans to exit OPEC, as well as a broader global turn toward economic nationalism, says Salar Ghahramani, a law professor at Penn State University and the president of Global Policy Advisors, a sovereign wealth fund advisory to investment banks and asset managers.
“An investing partner with deep pockets”
Even so, the shift does not signal an end to Saudi investment abroad. Technology deals and large-scale partnerships, including the Paramount-Warner deal, remain central to the country’s diversification strategy.
“If PIF, for instance, thinks that a brand-new deal with Tesla would be to its benefit, I think it would go that way,” Ghahramani says. “I think that generally, so long as their decisions are made by market-savvy individuals and not so much politically driven decisions, I think those global investment and partnership collaborations will continue.”
And since there’s no claim that PIF violated any deal with LIV in ending its support of the organization, the end of that arrangement is less likely to worry other businesses considering investment from the wealthy Saudi fund, Ghahramani says, noting, “All kinds of companies, in general, can always benefit from an investing partner with deep pockets.”
Carleton University’s Ettinger suggests that tech deals also don’t risk the same kind of negative press that followed the rise of Saudi-backed LIV, as Saudi Arabia positions itself to be “the kind of stable centerpiece of the Middle East” when the Iran war ends.
For PIF’s current focus, the deal with Five Iron may prove to be a better-suited investment than a continued role in LIV. The much smaller price tag, of course, may also make the deal easier to continue. And Five Iron’s Solomon argues the company can help expand access to golf within Saudi Arabia while offering a viable commercial return.
“We’re laser-focused on making sure it’s a good investment for them—making sure that the people like it and it actually does what they want to do, which is hit their goals of growing the sport of golf,” he says. “And if it can do that, I have no doubt that they will live up to their word and keep expanding things.”
Composite 25Trend 10
AI power users are pulling away from everyone else, Microsoft says
Artificial intelligence is helping knowledge workers do things that weren’t previously possible, according to a new report from Microsoft.
In the company’s 2026 Work Trend Index report, which includes results from a survey of 20,000 knowledge workers who use AI at work, 66% of the AI users surveyed say that AI allows them to spend more time on high-value work, and 58% reveal that they’re producing work they couldn’t have produced just one year ago. That number rises to 80% among a category of AI power users Microsoft dubs “frontier professionals.”
“Instead of just automating away what people used to do, and that’s an efficiency gain, what we’re seeing is much more exciting,” says Katy George, corporate vice president of workforce transformation at Microsoft. “What we’re calling ‘capability add.’”
Examples range from new uses of AI to find and address software security vulnerabilities, to salespeople being able to quickly get up to speed before a customer meeting to an extent previously impossible or impractical. And while the benefits may come first to those AI power users, they can, in turn, pass their knowledge on to colleagues—or benefit from a supportive environment: A previous Microsoft study found employees more likely to get value out of agentic AI when their managers serve as role models for effective use.
Those AI power users aren’t simply deferring to AI in every circumstance. George says some will, in fact, sometimes take longer to complete a particular task so that they can see how it can best be handled with the help of AI. But the survey also indicates that 43% of frontier professionals—and 30% of AI users overall—purposely do some tasks without AI assistance to keep their skills sharp. And 53% of frontier professionals—and 33% of AI users overall—intentionally take time before starting a task to decide what aspects should be done by AI versus a human.
That’s likely because human judgment, critical thinking, and the ability to provide quality control for AI results remain important, even when AI use becomes more prevalent. Anecdotally, reports of AI hallucinations and errors are well known, even as the technology improves for many tasks, so it’s not surprising that 86% of those surveyed by Microsoft say they tend to treat AI output as a starting point rather than a final answer.
“What declines is the amount of tactical, step-by-step execution work humans do themselves,” wrote Jared Spataro, CMO for AI at Work at Microsoft, in a blog post. “And what rises is the need for humans to set direction, define standards, and evaluate outcomes.”
In essence, even workers who don’t supervise any humans are now putting management and delegation skills to work supervising AI, along with their subject-specific training that lets them devise tasks for AI and evaluate the answers.
“People with real judgment and expertise are driving kind of the most effective use of AI,” George says.
That can also involve IT and cybersecurity experts setting up permissions and environments for AI operations, which can also overlap with skills already used to manage people’s digital access. “IT becomes the control plane for agent operations, extending the same rigor already applied to people and applications so that scale does not come at the cost of visibility,” according to the Microsoft report.
Still, that’s not to say that AI won’t significantly change the way people work, including eliminating certain jobs, with overall industry predictions of AI-related job loss and job creation varying widely. “Some jobs will change,” the report acknowledges. “Some will go away. And many that don’t exist yet will emerge.”
But while the report suggests that AI has the potential to make some knowledge workers more efficient, it’s not necessarily the case that employers should be demanding a particular level of AI usage from employees. The appropriate usage will likely follow from employees’ expertise and experimentation with the tools, George suggests.
“Therefore, we’re not so worried about, ‘Did you use it twice a day or 10 times a day,’” she says.
Composite 25Trend 10
OpenAI’s trillion-dollar AI bet is a study in ‘riskmaxxing’
As successful as OpenAI has been since the launch of ChatGPT, the company is operating in an extraordinarily expensive and risky corner of tech, building frontier AI models at massive scale. Its future, even its survival, is far from certain.
OpenAI is burning billions on top-tier AI research talent, carefully curated training data, and increasingly scarce computing power. Footing the bill is a growing cap table of VC and strategic partners, all betting on outsize returns within a few years.
Compute is the biggest cost. AI companies must lock in capacity years—not months—in advance. Data centers take years to build and bring online. That forces companies to forecast demand far ahead, then scramble to generate enough revenue to cover those commitments. If they underestimate demand, they leave revenue on the table. If they overestimate, the consequences can be existential.
OpenAI’s rival, Anthropic, must make a similarly precarious bet, but has bet more conservatively. Anthropic CEO Dario Amodei described the challenge on a recent podcast with Dwarkesh Patel: “The curve I’m looking at is: We’ve had a 10x-a-year increase every year. At the beginning of this year, we’re looking at $10 billion in annualized revenue. . . . I could assume that the revenue will continue growing 10x a year [but] I can’t buy $1 trillion a year of compute in 2027. If I’m just off by a year in that rate of growth, or if the growth rate is 5x a year instead of 10x a year, then you go bankrupt.”
OpenAI, by contrast, is playing a riskier game. The company has committed more than $1 trillion to building new data centers and leasing compute from an array of partners, including Amazon Web Services, CoreWeave, MGX, Microsoft, Nvidia, Oracle, and Arm. Oracle alone locked in a $300 billion, five-year data center partnership with minimum commitments running at about $60 billion per year by 2027, according to a PitchBook analysis. OpenAI has also contracted roughly $250 billion in compute from Microsoft, and pays about $5 billion annually back to Microsoft through its Microsoft Azure revenue share, PitchBook estimates.
All of this spending hinges on how quickly OpenAI’s revenue grows. The company is generating about $25 billion in annualized revenue, according to PitchBook, a roughly 40-to-1 ratio of obligations to current revenue. If it misses key growth targets, it may struggle to cover its compute and data center bills.
The Wall Street Journal reported last week that OpenAI missed internal revenue and user targets in early 2026, with CFO Sarah Friar privately warning leaders that the company may not be able to fund its future computing contracts if growth slows. OpenAI did not dispute the reporting. Instead, CEO Sam Altman and Friar said in a joint statement that they are “totally aligned on buying as much compute as we can.”
And boy, are they. By PitchBook senior private company analyst Harrison Rolfes’s estimate, OpenAI’s cash losses could mount to nearly $74 billion in fiscal year 2028 before it has any realistic path to breaking even by 2030.
“The Wall Street Journal’s reporting that OpenAI missed multiple monthly revenue targets this year after losing enterprise and coding share to Anthropic and Gemini is exactly the scenario that makes this math dangerous,” Rolfes tells Fast Company. “Every revenue miss compounds against a fixed obligation ladder that doesn’t flex.”
If OpenAI had locked down a product that no one else could replicate, or one far ahead of competitors, it might have a defensible moat. That could help offset the risks of such aggressive expansion. However, many analysts see that moat as somewhat limited.
“It has become clear that frontier models are rapidly commoditizing. DeepSeek repeatedly makes this clear,” says Columbia Business School professor Daniel Keum. “Switching costs are minimal. The main exceptions are firms like Google and Microsoft, which can embed AI into existing ecosystems that are very difficult to replace, such as Gmail, Google Calendar, and Microsoft Office.”
OpenAI may have gotten an early lead in a market growing at an “exponential 10x” clip, Keum adds, but it hasn’t built much differentiation or strong lock-in, particularly with consumers. Anthropic, by contrast, could prove more durable given its focus on enterprise customers, where switching costs tend to be higher.
And yet OpenAI recently raised $122 billion at an $852 billion valuation, suggesting investors still believe in a relatively fast “AI takeoff,” where businesses broadly integrate AI into their operations. That anticipated shift is what OpenAI and its peers are spending so heavily to prepare for.
“The risk to OpenAI isn’t sudden collapse,” Rolfes tells Fast Company, “but more that because the obligation stack is this large and this locked in, every revenue miss shrinks your options faster than most people appreciate.”
Both OpenAI and Anthropic are expected to go public in the near future, which will offer a clearer view into their financial risk. For now, investor sentiment appears to be tilting toward Anthropic’s more measured approach. In early April, reports appeared that people were shunning OpenAI shares on special market investor sites and flocking to Anthropic shares. Does OpenAI know something about the “AI takeoff” that investors don’t?
Composite 25Trend 10
Bose is rebooting its smart speakers for the Sonos haters
Bose is rethinking its approach to smart speakers.
While the company has released plenty of Wi-Fi-connected speakers over the years, its new Lifestyle Ultra line is a strategic reset, with a new platform that Bose spent the last few years building. (The name is also a nod to Bose’s original Lifestyle systems from the 1990s.)
The new Bose offerings include a $299 standalone speaker, a $1,099 soundbar, and an $899 subwoofer, which can also be combined into a surround system. Raza Haider, Bose’s president of premium consumer audio, says these are the first of many speakers that it will launch on the new technology stack.
“It’s a completely brand new platform, where we ripped the guts out of the old technology infrastructure,” Haider says. “It’s given us a hardware and software stack on which we can build for the future.”
[Photo: Bose]
Minimally smart
The main thing to know about Bose’s Lifestyle Ultra speakers is that they delegate most of the smart features to other companies.
Unlike previous Bose speakers, for instance, the Lifestyle line won’t support music controls through Bose’s mobile app. If you want to launch music from a phone, you’ll have to use Apple AirPlay, Google Cast, or Spotify Connect.
Those third-party systems will also handle multi-room audio, as Bose is stripping away the SimpleSync system that it previously used to connect Bose speakers around the home. Bose’s own app will merely handle setup for stereo pairs or surround sound in a single room.
That’s a markedly different approach from rival Sonos, which supports AirPlay and Spotify Connect but still emphasizes its own remote control app and multi-room features. And while Sonos has built its own music-focused voice assistant, Bose is leaning on Alexa+ instead, with plans to support other voice agents over time.
Although Sonos’ approach allows for tighter integration—for instance, you can use voice commands to move music between speakers—it can also backfire. When the company rushed out an app overhaul in 2024 filled with bugs and feature regressions, the resulting backlash decimated revenues and prompted its CEO to step down.
In leaning more on third parties and de-emphasizing Bose’s own app, Haider says the company is just trying to meet customers where they are.
“We basically heard from our customers that they want the music where they listen to their music,” Haider says. “They don’t want to go from Spotify Connect or AirPlay or Google Home into another app.”
[Photo: Bose]
Not getting stranded
As someone who’s accumulated and been vexed by a variety of smart speakers from Sonos, Amazon, Google, and Apple, I can see the appeal in Bose’s platform-agnostic approach.
My Google Nest speakers only connect with other Google Cast speakers. My Alexa speakers only connect with other Alexa speakers. My Sonos Beam soundbar and Sonos Move speaker sync with each other via AirPlay or the much-maligned Sonos app, but they don’t work with Google’s or Amazon’s multi-room systems. So maybe a speaker like the Lifestyle Ultra is the answer.
If there’s a reason for concern, it’s that Bose has walked away from one of its smart speaker platforms before. This month, Bose is discontinuing the SoundTouch platform it launched in 2013, cutting off internet-based features and security updates. Users who invested hundreds or thousands of dollars in SoundTouch speakers felt burned by the decision. (The company initially planned to discontinue AirPlay and Spotify Connect support as well, but later backtracked.)
Haider argues that SoundTouch had a good run by internet-connected consumer tech standards, but he hopes the new system will last even longer. Despite the seemingly minimalist strategy, he says a lot of work went into building a modular tech stack with room to grow and adapt to future changes. If Amazon were to rewrite some aspects of Alexa, for instance, it’s now easier to integrate those changes without overhauling the entire system.
In other words, by stripping away what wasn’t working, Bose may be able to avoid some of the missteps that have made smart speakers such a mess in the first place.
“It’s a reset in terms of a new platform that is future-ready, interoperable with partners, and the most external-friendly platform out there,” Haider says.
Composite 25Trend 10
The robotics pioneer of Roomba fame is now developing this 4-legged AI-powered robot
The robotics pioneer who helped unleash the Roomba vacuum is now betting that you might one day replace your beloved dog or cat with a plush robot that follows you around your home and adapts to your daily habits.
Colin Angle unveiled a four-legged prototype of that artificial pet, called the Familiar, on Monday. Imagine a creature the size of a bulldog with doe-like eyes and bear cub ears and paws, extending itself into a greeting stretch that invites you to pat its touch-sensitive fake fur.
“We chose a form factor that’s not a human, not a dog, not a cat, because we wanted to steer away from all of those preconceptions,” said Angle, who leads the startup Familiar Machines & Magic and before that was longtime CEO of Roomba maker iRobot.
This kind of lifelike machine — powered by the latest artificial intelligence technology — would not have been possible when Angle co-founded iRobot in 1990 or launched the first Roomba in 2002.
It’s hardly the first effort to build a pet-like household robot. Japanese electronics giant Sony, for one, famously introduced a small plastic robotic dog called Aibo in the late 1990s and rebooted the concept in 2018. But Angle believes the Familiar achieves something that “simply hasn’t existed before.”
“The challenge is to make something that’s not a watch-me toy,” Angle said in an interview with The Associated Press. “This is about having something that you want to hug, you want to pet. When it’s happy, that makes you happy. And it is large enough or mobile enough to follow you to the kitchen or drag you off the couch and take a walk.”
Angle said the robot will make emotive, animal-like sounds but won’t talk. But, mimicking a real pet, it has audio input “ears” and an AI system that can understand and learn from what you say to it. It benefits from the advances in generative AI sparked by chatbots like ChatGPT and can gradually adapt its behavior as it learns from the people around it.
“I couldn’t have done this six months ago,” Angle said.
Angle led iRobot for a quarter century as it turned Roomba into the first widely adopted home robot. Intense competition, especially from China, later threatened its success. Angle stepped down as CEO and chairman in 2024 after Amazon dropped its plan to buy the struggling Massachusetts company.
Familiar Machines was born soon after and remained in “stealth” mode in Woburn, Massachusetts until Monday, when Angle brought one of his Familiar prototypes to New York for The Wall Street Journal’s Future of Everything conference.
It could take a while before Angle starts selling the machines, but one target demographic is retired people who are past the peak age of pet ownership.
“Not because people suddenly stop enjoying pets, but the fear and obligation of caring for them are such that people are very reluctant to get new pets at older ages,” Angle said.
While most robot engineers take inspiration from science fiction, the idea of a familiar has deep roots in folklore, from a witch’s cat and wizard’s owl to the animal companions in Philip Pullman’s “His Dark Materials” fantasy novels.
“It’s an archaic, ancient word,” Angle said. To his surprise, he could also trademark it.
Angle has pulled together a number of prominent robotics advisers, including Marc Raibert, a pioneer of robot locomotion who founded Boston Dynamics, maker of the four-legged Spot robot; and Cynthia Breazeal, who invented the robot head Kismet and later the tabletop speaker robot Jibo, early attempts at imbuing robots with social expressions.
Many researched together at the Massachusetts Institute of Technology and share skepticism for the current fad of sleek humanoid robots that are designed to walk and move around like people but can’t yet do much useful physical work.
One of those advisers is Maja Matarić, a computer science professor at the University of Southern California who 25 years ago co-founded the field of socially assistive robotics — with the aim of designing robots that could give people social and emotional support.
When she first saw Angle’s prototype, she said she “immediately got down on the ground near it and had to hug it and pet it, then started to play with it to see what it would do.”
That people perceive the robot as adorable and not creepy will be key. Matarić said decades of research into human-robot interactions have shown that a robot that is “cute, personalized and vulnerable is much more appealing and lovable than the alternative.” It could be particularly useful in nursing homes or providing emotional support for mental health, she said.
Matarić said AI advances have also made it easier to broaden the impact to the general population.
“Before generative AI, robots could not readily understand what people were saying,” she said.
—Matt O’Brien, AP Technology Writer
Composite 25Trend 10
This 20-minute digital spring cleaning checklist saves time and money
We tend to treat our digital lives like a basement that never runs out of square footage.
Thousands of unorganized files in your downloads folder, the monthly subscription for a project management tool you haven’t opened in a year, and a professional bio that still claims you’re passionate about trends that aren’t even trends anymore.
In nerdy circles, we talk a lot about technical debt, which is the cost of choosing an easy solution now instead of a better one that takes longer. We rarely talk about digital rot: the accumulation of digital debris that slowly drains your focus, your storage, and your bank account.
Clearing out the clutter is a tactical necessity. Take 20 minutes or so to tidy up.
Subscription audit
The first step requires looking at your actual credit card statement rather than just the summary page. You’re looking for zombie subscriptions that have been quietly billing you for months.
If you haven’t used a tool to complete a billable task in the last 30 days, you should consider ending the subscription. If you haven’t watched a show on one of the 47 streaming services you pay for, sever those ties.
Once that’s taken care of, a helpful approach moving forward is a one-in, one-out policy: If you want to try a new service, an old one must go first. You’ll be surprised how much mental bandwidth you regain when you aren’t paying for stuff you never use.
Search, don’t sort
Many of us have been conditioned to build elaborate folder structures that we never actually navigate. Most of this is just digital hoarding disguised as organization.
A more efficient tactic is to create one single archive folder for the current year and move everything from your desktop and downloads directly into it. It’s the digital equivalent of stuffing all your clutter in a closet.
Seriously, try it: make a folder called “2026” and put as much as you can into it. Over the coming days and weeks, you’ll pretty quickly find the difference between what you actually need and what’s not all that important.
Why? Because modern search indexing is significantly better than any manual filing system you could create. If you haven’t searched for a file in six months, you likely never will. Clearing your visual workspace reduces the cognitive load every time you open your laptop.
Brush up your bio
Your digital presence is often the first thing people see, yet it’s usually the most outdated part of your professional life. Open your public profiles and delete any buzzwords that no longer reflect the current state of the industry.
Focus on results rather than roles. Instead of listing yourself as an expert in a specific type of software, describe how you helped your team reclaim time by streamlining a specific workflow.
You should also ensure your photo actually looks like you. If your headshot is several years old, it creates a subtle trust barrier that’s hard to ignore.
Composite 25Trend 10
How Substack became the new book tour
It’s hard enough to publish a book, but getting people to buy it is an entirely different battle. As new platforms reshape how readers gather and interact online, authors are finding that sometimes platforms built to showcase writing can also double as powerful engines for discovery.
The most high-profile example so far might be Girls creator Lena Dunham, who bolstered the traditional press tour for her new memoir Famesick with interviews and features on the newsletter platform Substack.
In an interview with Arielle Swedback for her On Substack newsletter (which is published, of course, on Substack), Dunham made the case in blunt terms: “Someone I trust told me that, in book sales at least, every single Substack follower is the equivalent of many more Instagram or X followers … While I don’t have the actual numbers, that feels anecdotally true to me. There’s an appreciation of the written word that suffuses this whole place.”
While promoting her memoir, Dunham did interviews with a range of the platform’s newsletters, from Emilia Petrarca’s Shop Rat, which has 32,000 subscribers, to Emily Sundberg’s Feed Me, with more than 150,000 readers. To Dunham’s point, many of these newsletters are built around tightly defined audiences that tend to be more engaged than those on broader social platforms.
“It’s been really interesting to see how committed certain audiences are. I love that a newsletter with more followers but a less engaged audience doesn’t have the same value as someone with a tiny but rabid fan base,” Dunham added.
And while Dunham may be the latest high-profile convert, she’s hardly alone.
“Ten years ago the publishing industry’s center of gravity was the bookstore and the New York Times list,” Andrea Barzvi, an agent and president of Empire Literary, tells Fast Company. “Today, discovery has been outsourced to algorithms. And the publisher relies more heavily than ever on social media—whether it’s the author’s own platform, or the mere power of social media.”
Social media’s influence on book sales takes many forms, including the wildly popular TikTok community BookTok, which has driven major sales for titles like The Song of Achilles, It Ends With Us, and The Seven Husbands of Evelyn Hugo. But while those platforms often depend on algorithmic luck, Substack offers something more direct: a line of communication between author and reader.
Jenn Lueke, author of Don’t Think About Dinner, says Substack offers a rare level of reliability. “I know my subscribers will actually see my posts,” she says, noting that the consistency makes readers more likely to try her recipes and follow her guides.
For Lueke, Substack became a tool for building her own community, one that followed her work before the book even reached the market. “I think someone who enjoys reading a newsletter might be more likely to enjoy reading a book,” she says. “My strategy was to utilize all social platforms I had to promote the book in different ways, with my Substack home being the center of it all.”
Some experts say Substack’s rise fits into a longer arc in publishing, one shaped by the early wave of self-publishing tools like Amazon Kindle Direct Publishing and Smashwords in the late aughts. Those platforms opened the door for self-published authors, but didn’t solve the marketing problem.
“That lack of support required self-published authors to be resourceful,” says Kris Austin, CEO of the self-publishing platform Draft2Digital. “Major publishing houses have taken note of indie authors’ business savvy and their ability to create fervent fanbases who are eager to purchase. This has led to traditional publishers moving away from status quo marketing spend, like print advertising, and leaning into newer opportunities.”
Those opportunities now extend well beyond Substack, giving authors multiple ways to cultivate an audience before a book even hits shelves.
“Press tours are decentralized now,” says Bookshop.org CEO Andy Hunter. “Individual creators can have a much bigger impact than old-school media.”
Dunham’s approach reflects that shift, and judging by early sales figures, it’s already paying off in a big way.
Composite 25Trend 10
The Pentagon wants lasers. Can anyone build them fast enough?
This article is republished with permission from Laser Wars, a newsletter about military laser weapons and other futuristic defense technology.
The U.S. military has a message for America’s directed energy industry: it’s time to build.
In a written posture statement submitted to the House Armed Services Committee ahead of a hearing on the U.S. Defense Department’s fiscal year 2027 budget request on April 29, Secretary of Defense Pete Hegseth stated that the Pentagon plans on buying “tens to hundreds” of directed energy weapons like high-energy laser systems in the coming years—the beginning of what Hegseth dubbed a “strong and consistent demand signal” to the U.S. defense industrial base that, after years of producing just “a limited number of prototypes,” the U.S. military is deadly serious about fielding such capabilities at scale.
Here’s the relevant section from Hegseth’s posture statement:
Directed Energy (DE) weapons represent a transformative capability, yet the Defense Industrial Base (DIB) is currently postured to produce only a limited number of prototypes. There are significant vulnerabilities and gaps in our DE defense manufacturing capabilities. To address this, the Department must create a strong and consistent demand signal for the production of greater quantities of these weapons, on the order of tens to hundreds of units.
This increased demand is essential to enable the DIB’s manufacturing capacity to mature and scale to meet the tactical innovation of the warfighter. Overcoming the “business as usual” acquisition mindset is paramount. The Department must reform its procurement processes, warfighting tactics, and policy limitations to “demystify” Directed Energy weapons and facilitate their integration into the force structure. This includes developing new concepts of operation, training programs, and support infrastructure to ensure that these advanced weapons can be effectively fielded to our warfighters and employed on the battlefield.
The successful integration of Directed Energy weapons will require a concerted effort to overcome institutional inertia and embrace a new way of thinking about warfare. The Department’s commitment to creating a demand signal is the first and most critical step in this process.
While senior military and defense officials have vocally endorsed fielding directed energy weapons at scale in 36 months or installing “a laser on every ship,” Hegseth’s statement offers a more grounded (and familiar) diagnosis for observers of the U.S. military’s decades-long laser weapon ambitions: the technology has advanced, but the institutional mechanisms to transition mature systems to the field have not. The defense industrial base simply cannot invest in the manufacturing and supply chain capacity required for production at scale if it can’t predict how many systems it will actually be asked to build, especially if promising initiatives continually perish in the “valley of death” between research and development and procurement
The defense industry has been making this point for years. A January 2024 report from the National Defense Industrial Association (NDIA) trade group on directed energy weapon supply chains, which is on based in-depth research and interviews with dozens of key industry stakeholders and subject matter experts, found that the lack of a consistent demand signal “was raised many times by industry leaders as negatively impacting all levels of the supply chain.”
“Existing [directed energy weapon] supply chains can only produce small numbers of systems with long lead times,” the NDIA report says. “Once DoD’s strategic goals are articulated, appropriate DEW systems should be transitioned to programs of record and multi-year contracts used to send an extended demand signal. A clear, sustained demand signal, accompanied by the overarching strategic vision, will provide industry with the assurance that they can begin to make the internal investments necessary to secure DEW supply chains for the future.”
This assessment isn’t wrong. Despite ramping up laser weapon efforts following a deliberate shift from bulky chemical systems to more reliable, compact, and efficient solid-state and fiber laser technology in the 2000s, the last two decades have been marked by abandoned projects. Here are some recent examples:
U.S. Army officials touted its 50 kilowatt Stryker-mounted Directed Energy Maneuver-Short Range Air Defense (DE M-SHORAD) as a major breakthrough when it deployed to the Middle East for real-world operational testing in 2024, but the U.S. Government Accountability Office (GAO) concluded the system “was not mature enough” to transition to a program of record.
Army officials told Congressional Research Service as recently as this past January that they planned on transitioning the ambitious cruise missile-killing 300 kw Indirect Fire Protection Capability-High Energy Laser (IFPC-HEL) weapon to a program of record, but now say they only plan on taking delivery of a single system to use as a testbed to inform future laser weapon development efforts.
The U.S. Navy’s 60 kw High Energy Laser with Integrated Optical Dazzler and Surveillance (HELIOS) weapon system, which only recently began testing at full power and frying drones aboard Arleigh Burke-class guided missile destroyer USS Preble after years of delays, has effectively disappeared from the service’s fiscal year 2027 budget request outside of a handful of sustainment dollars.
The U.S. Marine Corps returned its five Compact Laser Weapon System (CLaWS) units to Boeing in pursuit of a “more deliberate programs of record,” years after touting the system as “the first ground-based laser approved by the Department of Defense for use by warfighters on the ground” (and without any explicit funding for laser weapon R&D in its fiscal year 2027 budget request).
The U.S. Air Force spent years experimenting with Raytheon’s High-Energy Laser Weapon System (HELWS) for counter-drone missions but abandoned the effort without successfully transitioning the system to a program of record, although the service appears poised to once again pursue ground-based laser weapons for airbase defense.
These failures share a common pattern, according to a detailed 2023 GAO report on the Pentagon’s directed energy weapons efforts: projects advanced through prototyping without ever securing formal transition partners or drafting agreements that would bind developers and the acquisition community to shared requirements, timelines, and funding responsibilities. The Navy’s HELIOS effort, for example, identified a notional transition partner but never documented agreements detailing how to resolve various power and cooling integration challenges before the system headed to an actual warship for installation.
The Air Force’s HELWS spent more than three years in development before the service even identified a transition partner, and when it did, the relevant program office had neither the funding nor the mandate to take it on. The Army’s comparatively more disciplined approach—embedding transition teams in prototyping efforts, drafting early capabilities documents, and regularly convening stakeholders to plan for future doctrine, training, and maintenance—shows what the other services didn’t do, and even that wasn’t enough to save DE M-SHORAD from demilitarization. There is simply too much “institutional inertia,” as Hegseth put it, to allow promising systems to drift toward obsolescence rather than fight the bureaucratic battles required to turn them into programs of record.
So what does a “clear, sustained demand signal” actually look like? The Pentagon’s fiscal year 2027 budget request contains a few elements that indicate the beginnings of a firm institutional commitment to fielding laser weapons (although, as one defense official recently reminded me, justification books rarely survive contact with the budget process).
First, the Joint Laser Weapon System (JLWS): a containerized 150-300 kw laser weapon designed to defeat incoming cruise missile threats as part of the Trump administration’s new “Golden Dome for America” missile defense shield. As I’ve previously reported, the fiscal year 2027 budget documents lay out a planned R&D investment of $675.93 million through fiscal year 2031 to develop the joint Army-Navy system based on lessons from HELIOS and IFPC-HEL, among other higher-power laser weapon efforts. And while there are no explicit procurement plans yet, this investment will likely be augmented by additional funds from the $452 million the Pentagon has requested specifically for directed energy weapons as part of Golden Dome separate from the services.
Second, the Enduring High Energy Laser (E-HEL): the modular 30 kw laser weapon the Army envisions as its counter-drone system of choice and eventual program of record. Beyond ongoing directed energy R&D efforts, the service has stated that it plans to “produce and rapidly field” 24 E-HEL systems over a five-year period, with plans to purchase two at a time for roughly $17 million apiece for the first two years before subsequently ramping up to batches of five. This program appears to be moving faster than most laser efforts before it, with the first E-HEL prototype expected no later than the second quarter of fiscal year 2026 and initial procurement units slated for delivery by end of fiscal year 2027. Even the Navy is exploring the E-HEL’s potential naval applications, per the service’s fiscal year 2027 budget request.
It’s also worth noting that Hegseth’s posture statement invokes the 23 new Portfolio Acquisition Executives (PAE) that the Pentagon has already established across the services, which are designed to transform the U.S. military acquisition processes to “prioritize performance and accountability.” A dedicated directed energy PAE with real budget authority behind it could prove a concrete test of whether this new framework changes outcomes rather than just incentive structures (although the posture statement doesn’t explicitly commit to one).
Are bold declarations from military and defense leaders, a massive R&D budget, and renewed promises of programs of record a strong enough directed energy demand signal for the defense industrial base? Recent laser industry moves, both domestic and international, suggest as much. Huntington Ingalls Industries announced a new laser integration and test facility in support of the E-HEL effort in September 2025. The following November, IPG Photonics announced the grand opening of a new manufacturing facility in Huntsville, Alabama, dedicated to developing and producing laser weapons for defense applications.
In January, nLight announced a 50,000-square-foot laser weapon manufacturing addition in Colorado before unveiling an expansion of Italy operations to support European directed energy development in April. Australia’s Electro Optic Systems (EOS) opened a laser weapon production hub in Singapore in February amid ongoing discussions with the U.S. and other potential customers. AV, the maker of the LOCUST Laser Weapon System that has become a fixture of U.S. counter-drone operations, announced a $30 million manufacturing expansion in Albuquerque, New Mexico earlier in March. Finally, start-up Aurelius Systems announced a brand new division focused on building fiber laser source modules in the U.S. in late April.
But manufacturing expansions alone aren’t enough for the U.S. military to meet its near-term goal of rapidly fielding directed energy weapons at scale. Part of the problem is that laser weapons are arguably more complex and time-consuming to produce than, say, Raytheon’s Coyote interceptors; the new EOS Singapore hub, for example, can only produce five to 10 laser weapon systems annually, per company executives. But more importantly, a demand signal hundreds of laser weapons is only meaningful if the entire directed energy supply chain is ready to answer the call—and according to the NDIA report, it is far from ready.
First, many critical components in laser weapons currently face long lead times due to lack of capacity. As the NDIA report notes, the precision mirrors and lenses that shape and direct laser beams require highly specialized grinding and polishing to tolerances that can take 12 to 18 months to produce for a single large optic. Beam directors, the devices responsible for precisely aiming and controlling the laser beam, are built by just two or three companies in the U.S., with lead times that regularly stretch beyond two years. Adaptive optics, which compensate for atmospheric distortion in real time, have only two or three suppliers for non-medical applications, with lead times of 18 to 24 months.
Specialized optical fibers essential for efficient energy transmission are so niche that one NDIA interviewee mentioned a Scandinavian company as among the few viable suppliers. Ceramic laser gain materials are sourced from a single company in Japan. Diffraction gratings critical to laser amplification come from a single industry supplier. Beam dumps used in testing—a component so routine it barely registers in program discussions—are manufactured exclusively by one company in Israel, with lead times that have stretched to a year.
Second, the raw materials required to make these components are subject to their own geopolitical bottlenecks, as I’ve previously noted. The essential lasing medium in most high-energy laser weapons is a solid-state or fiber gain medium doped with rare earth elements—neodymium, erbium, thulium, ytterbium. Unfortunately, Chinese exports accounted for 74% of U.S. rare earth element imports between 2018 and 2021, while Beijing controls more than 85% of global processing capacity. The laser diode pumps that drive most solid-state laser systems are typically built from gallium arsenide, but China controls 98% of global gallium production and announced fresh export controls in the summer of 2023.
Germanium, a primary material in the infrared optics, is similarly exposed: 54% of U.S. imports come from China and are subject to those 2023 export controls. Even the copper used in laser weapon thermal management systems runs through Chinese processing, with 41% of all refined copper originating in China as of 2022 despite the US’s substantial domestic ore production. A determined U.S. effort to scale military laser weapon production to hundreds of units would face Beijing-controlled chokepoints at almost every major component layer.
There’s a third constraint lurking beneath the manufacturing and materials challenges: the U.S. simply does not have enough people trained to build laser weapons at scale. The NDIA report identified three specific workforce categories facing acute shortages in the directed energy sector: optical coatings specialists, power electronics engineers, and opto-mechanical engineers. Optical coating construction and application is, in the words of industry participants, an “artform” that takes years to master, and there are only a handful of U.S. companies devoted to defense-grade coatings.
Optics graduates are also scarce: only a handful of schools in the U.S. have dedicated optics programs, and they face intense competition from the medical device and consumer optics industries, which pay better and don’t require security clearances. Finally, power electronics engineers with the unique experience needed for the power conversion and charging systems in directed energy weapons are increasingly hard to find as broader demand for electrification across commercial industries drains the same talent pool.
Taken together, these challenges are the reason “tens to hundreds” of directed energy weapons has remained an aspirational goal rather than a reality for so long—and they won’t be solved by a demand signal alone. A long-term solution will require sustained, coordinated investment across manufacturing, materials, and workforce development. While the Pentagon is already pursuing potential solutions with $100 billion requested in fiscal year 2027 to “supercharge” the defense industrial base, those investments will take years to translate into consistent production capacity,
Still, Hegseth’s posture statement represents the clearest and most senior articulation yet that the Pentagon understands the systemic problems that have held back its directed energy programs and intends on addressing them. Whether it’s also a sufficient step will depend on whether E-HEL’s transition to a program of record actually happens on schedule, the $675 million JLWS investment survives the budget process, and the supply chain and workforce investments needed to back up a demand signal for hundreds of systems materialize alongside it.
The defense industrial base has heard this kind of rhetoric before. What it needs now is the multi-year contracts, programs of record, and upstream investments in materials and workforce that will empower it to actually respond. The next budget cycle will reveal whether this time is different.
This article is republished with permission from Laser Wars, a newsletter about military laser weapons and other futuristic defense technology.
Composite 25Trend 10
iPhone owners could get up to $95 after Apple settles AI lawsuit for $250 million
Owners of some iPhones are in line to get cash payments of up to $95 from Apple after the company on Tuesday reached a $250 million settlement in a class-action lawsuit for false advertising of its artificial intelligence capabilities.
Apple trumpeted new AI features for its virtual assistant Siri when it rolled out the iPhone 16 in 2024, part of new software updates that the company billed as “Apple Intelligence.”
The company has been scrambling to keep up with tech rivals amid the AI boom but still hasn’t delivered on the Siri revamp two years later.
The lawsuit, filed on behalf of U.S. consumers in the San Francisco federal court for the Northern District of California, alleged that Apple deceived consumers with a marketing campaign that promoted features that did not yet exist and misled them into buying the devices.
Lawyers for the iPhone buyers asked a court for preliminary approval of the proposed $250 million settlement, according to a court filling. If approved by a judge, it would be one of the biggest ever for Apple.
The settlement covers about 37 million devices bought in the United States between June 10, 2024 and March 29, 2025, including all iPhone 16 models and the iPhone 15 Pro and iPhone 15 Pro Max.
Owners are eligible for a payment of at least $25 for each device, and that amount could go up to $95 depending on how many other claims are filed “and other factors,” the filing said.
Customers will be notified by email or mail that they can file a claim on a settlement website, it said.
Apple, based in Cupertino, California, was caught off-guard by the intense consumer interest in the Siri AI features. Buyers were angered after finding out that the new features would be released later than expected, the filing said.
They “would not have purchased the Eligible Devices or would have paid significantly less, had they known Enhanced Siri features were not available,” the filling said.
Apple’s AI features remain in development even as rivals Google and Samsung have been rolling out more of the technology on their own devices. The company is expected to unveil its Siri upgrade this year, most likely at its annual developer conference next month.
—Kelvin Chan, AP Business Writer
Composite 25Trend 10
AI labels were supposed to help users spot fakes. Here’s why they’re failing
Fake accounts have been around as long as social media. So when it was recently revealed that a “hot girl” MAGA personality named Emily Hart was actually a 22-year-old male medical student in India, it might have seemed a little mundane. Just another catfisher, another sock puppet, another scammer—the internet is full of them.
Except this one had photos. And videos. And thousands of followers across multiple networks with some posts getting millions of views. Emily Hart was a full-on influencer, not just some anonymous egg. The person who created Emily confessed to Wired that while the account was active, he was making thousands of dollars every month from posting softcore videos to an OnlyFans competitor and merchandising.
Emily’s creator is not a developer. He’s just a cash-strapped student with a good sense of American political culture and a Google Gemini account. But the curious case of Emily Hart has exposed how AI has made it incredibly easy for almost anyone to create convincing content and game the system of engagement on social media.
{"blockType":"mv-promo-block","data":{"imageDesktopUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/03\/media-copilot.png","imageMobileUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/03\/fe289316-bc4f-44ef-96bf-148b3d8578c1_1440x1440.png","eyebrow":"","headline":"\u003Cstrong\u003ESubscribe to The Media Copilot\u003C\/strong\u003E","dek":"Want more about how AI is changing media? Never miss an update from Pete Pachal by signing up for The Media Copilot. To learn more visit \u003Ca href=\u0022https:\/\/mediacopilot.substack.com\/\u0022\u003Emediacopilot.substack.com\u003C\/a\u003E","subhed":"","description":"","ctaText":"SIGN UP","ctaUrl":"https:\/\/mediacopilot.substack.com\/","theme":{"bg":"#f5f5f5","text":"#000000","eyebrow":"#9aa2aa","subhed":"#ffffff","buttonBg":"#000000","buttonHoverBg":"#3b3f46","buttonText":"#ffffff"},"imageDesktopId":91453847,"imageMobileId":91453848,"shareable":false,"slug":"","wpCssClasses":""}}
It also raises the question: Is anyone looking out for us out there? How can you tell what’s real and what’s not anymore? And who is responsible for alerting social media users that the images they’re looking at might have come from AI?
The fake influencer template
The major implication of the story isn’t about a single AI influencer. It’s that this is the tip of the iceberg. AI has made creating online personas like Emily so easy that it’s enabled deception at scale. The Wired story points to other pro-Trump fake influencers like Jessica Foster, but you don’t have to look very far in your Instagram Explore page before you spot something AI-generated, and it’s rarely disclosed. The Emily Hart case proves that the template is cheap, fast, lucrative, and easy to copy.
All the major social networks have policies governing AI content. While they vary in detail, the gist is generally the same: Synthetic images must be disclosed—especially if it could be construed as real and the subject matter involves sensitive subjects like politics, health, finance, and current news. If the account doesn’t identify AI content, it could be frozen, demonetized, or banned.
But those penalties exist almost entirely on paper. In practice, enforcement is difficult, partly because detecting AI content is getting more difficult by the day. Most state-of-the-art image generators are light-years ahead of the models that created the first “Will Smith eating spaghetti” video, and telltale artifacts like extra fingers and disappearing background characters have largely become a thing of the past. Without watermarks, even automated systems have a difficult time parsing AI images from real ones just by looking at them.
The ‘nutrition label’ that keeps getting lost
A new standard was supposed to fix this. Content Credentials are a way to track how an image was created and modified throughout its life cycle. That information can be preserved in the image’s metadata, so the site displaying it can more easily tell whether it’s AI-generated, potentially passing on a label or warning to the user. The idea is that, as you scroll your social feed, any image would have a tiny icon next to it that would reveal its history when clicked.
However, even though this technology has existed for years and ostensibly has the support of major tech companies such as Adobe, Google, and Nvidia, social platforms haven’t adopted it consistently. Seeing the label is rare, and a Washington Post report found that social networks often strip out the metadata that enables Content Credentials. This isn’t necessarily nefarious—it follows a best practice from the early days of the web when every byte was precious. But the fact that it’s still happening shows there is little enthusiasm to make the system work.
Would a label make any difference? Emily’s creator says he believes many of his followers didn’t care whether the images he was posting were AI or not. That may be true for some, but data suggest labels can alter people’s propensity to engage with AI content. A 2024 study found that labels on AI-manipulated media reduced belief in the claims. The study also found that wording matters: “manipulated” or “false” were more impactful than process-based labels alone.
In other words, labels help, but weak labels help weakly. A buried “AI info” tag is not the same as a clear warning that an image might depict a person who does not exist.
Platforms like Facebook, Instagram, YouTube, and TikTok already process and modify content at scale. They’ve spent two decades building the art of detecting copyright violations, nudity, spam, and engagement signals. It is hard to believe they are incapable of building a clearer label for AI-generated people.
It’s the incentives, stupid
So why don’t they? The uncomfortable answer is that the incentives point the other way. While platforms want to keep bad content out, they are more motivated to keep people posting, scrolling, sharing, and buying. AI-generated material fits neatly into that machine because it is cheap to make, easy to personalize and highly compatible with engagement-driven feeds.
Mark Zuckerberg has been unusually direct about this, describing AI-generated material as “a whole new category of content” that he sees as important for Facebook, Instagram and Threads. That doesn’t mean Meta or any other platform wants deception (which, again, is a subcategory of AI content). But it does mean the companies have a business reason to welcome more synthetic content, and making the labels too strong or too visible could dampen the engagement they’re trying to encourage.
The calculus could change, though. Europe’s AI Act includes transparency obligations for deepfakes and certain AI-generated public-interest content, with related rules taking effect this year. Should platforms start to rack up major fines for poor labeling, things could change in a hurry. Advertiser pressure would help, too, since appearing next to deceptive content is bad for business. Finally, and crucially, there’s audience behavior: if users begin to feel like they can’t trust what they’re seeing on a network, they might, over time, stop engaging with that network.
The burden has shifted
Right now, the responsibility for detecting AI content is falling largely on the user, with the social platforms not prioritizing the technical progress that might help, and regulators only beginning to act. And you might question what’s the point—many of Emily’s followers no doubt knew she was virtual but followed, engaged, and maybe even forked over some money anyway. However, that choice—to engage or not with a virtual influencer is robbed from you if you don’t know it’s virtual in the first place.
The technology industry has spent years presenting provenance as a central answer to synthetic media. Adobe, Microsoft, Meta, OpenAI, Google and others have backed standards, joined coalitions, made public commitments and embedded Content Credentials into their tools. Fine. Then show it to people. Make it visible before the share, before the follow, before the subscription, before the merchandise purchase. Because if the only way to learn that an influencer is fake is to wait for a magazine investigation, the disclosure system has already failed.
{"blockType":"mv-promo-block","data":{"imageDesktopUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/03\/media-copilot.png","imageMobileUrl":"https:\/\/images.fastcompany.com\/image\/upload\/f_webp,q_auto,c_fit\/wp-cms-2\/2025\/03\/fe289316-bc4f-44ef-96bf-148b3d8578c1_1440x1440.png","eyebrow":"","headline":"\u003Cstrong\u003ESubscribe to The Media Copilot\u003C\/strong\u003E","dek":"Want more about how AI is changing media? Never miss an update from Pete Pachal by signing up for The Media Copilot. To learn more visit \u003Ca href=\u0022https:\/\/mediacopilot.substack.com\/\u0022\u003Emediacopilot.substack.com\u003C\/a\u003E","subhed":"","description":"","ctaText":"SIGN UP","ctaUrl":"https:\/\/mediacopilot.substack.com\/","theme":{"bg":"#f5f5f5","text":"#000000","eyebrow":"#9aa2aa","subhed":"#ffffff","buttonBg":"#000000","buttonHoverBg":"#3b3f46","buttonText":"#ffffff"},"imageDesktopId":91453847,"imageMobileId":91453848,"shareable":false,"slug":"","wpCssClasses":""}}
Composite 25Trend 10
Why China’s feverish use of AI tools could shape how the tech is used globally
On a recent weekday, around 50 people gathered outside the headquarters of a Chinese mobile internet company, waiting to get help with installing an artificial intelligence assistant.
The scene in Beijing, China’s capital, was repeated for days at several events and was also seen in the southern technology hub Shenzhen in March, as engineers helped crowds trying to set up the popular AI “agent” OpenClaw on their laptops.
“I’m worried about falling behind in technological developments,” said Sun Lei, a 41-year-old human resources manager at the Cheetah event. She said she hoped the tool might help her source and screen resumes across various recruitment platforms.
More than a year after OpenAI’s Chinese rival DeepSeek stunned the world with its advanced AI model, China has become a testing ground for mass use of AI tools. AI models built in the United States still dominate in raw computing firepower, but Chinese people and businesses have rapidly embraced the technology, facilitating its swift and widespread adoption in almost every possible field.
As global AI adoption rises quickly at workplaces and in daily lives, ordinary Chinese are using AI for all sorts of things, from booking and planning travel, ordering food and hailing rides. Of its 1.4 billion population, more than 600 million were using generative AI as of December, a 142% increase from a year earlier, according to a report by the government-controlled China Internet Network Information Center.
And, with the recent surge in the use of “agentic” AI like OpenClaw including for many Chinese businesses, the consumption of data by AI models has also risen. Measured in what computer scientists call tokens, or units of data such as part of a word, the weekly share used by Chinese AI models has recently surpassed U.S. models, according to OpenRouter, an AI “gateway platform” that tracks data and enforces security across different AI models.
AI adoption positions China as a ‘world leader’
Jason Tong, a 64-year-old retiree in Shanghai who has worked as an IT engineer, has been using AI chatbots such as Doubao and Kimi for everyday queries since they were first introduced a few years ago.
He began paying closer attention to his health and in early March joined a blood glucose monitoring service run by a Shanghai-based company that uses an AI model to generate tailored health advice. He has found its personalized, rapid responses helpful.
Widespread adoption of AI applications in everyday life is inevitable, Tong believes, “Just as carriages were eventually replaced by trains, this is bound to happen.”
Chinese products incorporating AI such as cars and robots are making major advancements, from humanoid robots with advanced cognitive capabilities to AI systems that drivers can use for more complicated tasks like making a restaurant reservation.
“The (AI) competition is clearly shifting from models to ecosystems,” said Lizzi Lee, a fellow at the Asia Society Policy Institute’s Center for China Analysis focused on economics and technology. “Chinese users are basically acting as real-time testers at scale.”
Chinese technology companies like Tencent, Alibaba and Baidu are also racing to commercialize AI. Tencent integrated OpenClaw into WeChat, China’s own “super-app” which is primarily a messaging tool but can also be used to do things like ordering food and making payments. Alibaba is embedding “agentic” AI into its workflows.
OpenClaw fuels wider use of China AI applications
OpenClaw, originally created by Austrian software developer Peter Steinberger last year, won quick and enthusiastic use thanks to its ability to use various tools to complete complicated tasks.
Zhao Yikang, a Chinese college student in Macao, uses OpenClaw in both his studies and daily life.
He was struck by how low-cost and efficient it is, using it to automatically generate promotional videos and manage social media accounts during his internship at a real estate agency in the southern Chinese city of Zhuhai.
“AI can understand things in a second,” Zhao said. “You just need to act as a commander and tell it what to do.”
Preparing to start a photo services business after graduation, Zhao asked AI to build a company website. Within 10 minutes, it had generated a fully functional site for less than 5 yuan (70 cents).
At one point, Chinese authorities issued several warnings about potential security risks over OpenClaw AI “agents” like data leaks as installations spiked, the broad interest had not faded.
Chinese companies increasingly are setting internal targets for boosting use of AI to improve efficiency, said Janet Tang, a partner & managing director focused on technology at consultancy AlixPartners.
There are “a lot of application scenarios,” said Wang Xiaogang, co-founder of the Chinese AI software company SenseTime and chairman of ACE Robotics. “The industry is developing very fast and the people, they are very open and they’re eager to try the AI in a lot of scenarios.”
US export controls both help and hinder AI use in China
China has sought to stack the deck in its favor, investing heavily in nurturing talent and ensuring access to abundant, affordable electricity for power-hungry AI developments and breakthroughs.
To achieve technology breakthroughs including in AI, Chinese leaders have pledged an annual average growth of at least 7% in nationwide spending on research and development in the country’s five-year plan until 2030. An “AI plus” national blueprint outlines steps to integrate AI into many areas of life, from healthcare to education. Judges in Shenzhen processed 50% more cases last year, a court said, partly with the help of an AI tool assisting judicial processes.
However, limited access to the some of the world’s most advanced computer chips due to U.S. restrictions remains a bottleneck for China’s AI advancement.
“Export controls on tools have slowed China’s chipmaking capabilities, and are the Achilles’ heel of many AI labs that need advanced AI chips,” said Samm Sacks, a senior fellow at New America focused on Chinese technology policies.
But the controls also have led to improved coordination of design, manufacturing and adoption across China’s tech supply chain. “Over time this dynamic could fuel, not foil, China’s ambitions,” Sacks said.
China is becoming an AI ‘innovator’
When China’s DeepSeek released its long-anticipated V4 AI model preview last month, one major change was that it’s supported in part by computer chips made by Chinese tech giant Huawei. That means less dependence on top U.S. chipmakers such as Nvidia.
A recent report by Stanford University’s Institute for Human-Centered AI says the U.S.-China gap in top AI models’ performance has “effectively closed.”
U.S. policymakers and top AI firms including Anthropic and OpenAI have accused Chinese AI startups of stealing U.S. AI technologies. China says such allegations are groundless.
Lian Jye Su, a chief analyst at the research and advisory group Omdia, believes any AI gap between the U.S. and China will continue to narrow, despite U.S. export controls and China’s Great Firewall, the ruling Communist Party’s massive internet filter and censorship system.
Analysts including Su believe that hurdles such as the Great Firewall are will likely impacts China’s AI use in limited ways, given that the technology already is being tested, integrated and scaled up under China’s controlled internet environment.
“It won’t be long before China moves from fast follower to parallel innovator,” he said.
AP researcher Shihuan Chen and journalists Dake Kang and Matt O’Brien contributed.
—Chan Ho-Him, AP Business Writer
Composite 25Trend 10
How a Texas vegan cheese-maker used Claude and Manus to fight back against a big shipping company
AI isn’t all about automating core business functions at Fortune 500 companies. Small and medium-sized businesses can also use AI to optimize, economize, and in some cases compete more effectively against much larger rivals.
An Austin, Texas-based vegan cheese-maker called Rebel Cheese used it to level the playing field against a larger supplier. Specifically, the company developed a small system of AI tools to help it claw back overcharges from a major shipping carrier.
The company is perhaps best known for winning a $750,000 investment from Mark Cuban, money it used to grow Rebel Cheese into what it says is now a $20 million business. Cuban recently spoke about the company’s crafty use of AI on stage at the Convergence AI event in Dallas.
The Problem
Rebel Cheese ships tens of thousands of orders of perishable, handcrafted vegan cheese across the country. The holiday season is by far its busiest period. “Q4 is all hands, heads down, get it out the door, make sure customers are happy,” the company’s cofounder, Kirsten Maitland, wrote in a recent blog post. “There’s no time to stop and analyze anything.”
After this past holiday season, Maitland took a look at the company’s bank account, and something seemed off. Rebel Cheese had just had its best holiday season ever, yet the numbers didn’t reflect it. So she started digging to find out where the profits were leaking away. She discovered that the company had paid $250,000 more for shipping than planned.
Hiring new employees to research and fix the problem wasn’t in the cards. So Maitland turned to Anthropic’s Claude. “I handed it a year of invoices and a contract,” she tells Fast Company in an email exchange, “and it found patterns I would have needed a forensic accountant to surface, which would have been time-consuming and expensive.”
She says the carrier’s shipping invoices run hundreds of pages per week, with fees layered inside fees. “Most shippers don’t have the time or tools to audit them,” Maitland says. For the carrier, the complexity was not a bug but a feature, and a profitable one.
Her analysis turned up several causes, not just one. “Some were our fault, like significant weight overages on our packages, which we could fix,” she says. “The rest were on the carrier: They had put a custom contract in place for us. Under that contract, any package bulging or weight overage triggered drastic price spikes.”
By the time Maitland sat down with representatives from the shipping company, she had analyzed a year’s worth of data and could show them exactly which contract clauses were doing the damage. The biggest overcharges mapped to a new weight limit the carrier had implemented, but not communicated, in early 2025. Their response was: “Well, you should have caught it.” She vowed never to let that happen again.
The build
To build the actual programs that read the invoices and request refunds, Maitland used Manus, an AI orchestration layer that coordinates work among various agents and subagents, using different models for different tasks. Maitland says she also tested Bolt, Lovable, and Relay, but found that Manus handled the job more easily and accurately.
After a lot of experimentation and discovery, she was able to architect a system that automated a data-heavy auditing process that had previously required manual review of tens of thousands of shipments. The build unfolded in four distinct phases:
1. Standardizing the “Truth.” The process began with data preparation. Maitland created two simple comma-separated value (CSV) templates. A “Zone Data File” contained Rebel Cheese’s negotiated contract rates, while a “Transaction File” contained weekly invoice data. This gave the AI a clear structure for comparing “what we should pay” against “what we were actually billed.”
2. Designing the Blueprint. She uploaded example invoices, Rebel Cheese’s carrier contract, and a presentation detailing the results of her Claude-assisted investigation into the overcharges to Manus. Rather than immediately asking the AI to “build a tool,” Maitland first used it to generate a comprehensive “Requirements and Design Document.” The document served as a technical blueprint, laying out the business logic for “fuzzy weight matching” and methods for flagging discrepancies. The step ensured the AI understood edge cases like fuel surcharges and weight brackets before a single line of code was written.
3. Building via Orchestration. Maitland then asked Manus to build a tool based on the blueprint document. Her prompt began: “I need you to build a standalone, single-page web application that acts as a Carrier Billing Discrepancy Detection Tool (works for any carrier — UPS, FedEx, USPS, or your specific shipping partner).” She stipulated that the tool should flag every shipment where the actual charge exceeded the contracted rate by more than ten cents. Those overcharges would then be sent to the carrier alongside a request for credit for every discrepancy it could not justify.
4. Continuous Auditing and Strategic Insights. Once the tool flags overcharges, Maitland feeds the data back into Claude, which analyzes the logs for higher-level patterns, such as shipping zones where costs are spiking. That transformed the tool from a simple invoice checker into a permanent recovery system.
The Result
The system Maitland built audits all shipping charges by comparing carrier invoices against Rebel Cheese’s contracted rates. It flags every discrepancy and generates a report that’s sent directly to the shipping carrier, requesting credit for every overcharge that can’t be justified. Maitland says the carrier has approved and credited every claim the system has submitted so far. She pays about $200 per month for her Claude and Manus subscriptions, and says the company is now saving between $1,500 and $4,000 every week.
Now that Rebel Cheese has gained experience with, and trust in, AI automation, the company is already using the technology for other core business functions, Maitland says. She built an agent that monitors the fundraising pipeline, researches VCs, and prepares her for investor meetings. She also built a site that handles inbound donations and partnership requests. Another tool uses historical data to draft ads.
“The bigger shift was realizing this is what closes the gap for companies our size,” she says. “We don’t have an engineering team. We don’t have a data analytics team. A few years ago, I would have had to hire a consultant . . . now I can do the work myself in an afternoon.”
Composite 25Trend 10
AI? No thank you! 3 truly free, no-AI apps for the overwhelmed
Is it just me or is every app update lately promising to “reimagine my workflow” with a new generative assistant? My toaster probably has a chatbot now.
We’ve reached a point where software is trying so hard to think for us that it’s actually making it harder to just do the work. In other words, when everything is “smart,” everything is noisy.
If you’re feeling the same AI fatigue while trying to manage a career, a household, and a few side projects, here are three pure-utility apps that are actually free and refreshingly, wonderfully dumb.
Joplin
If you’ve been in the tech world for a while, you remember when Evernote was the king of the mountain, before it became a bloated, expensive quagmire.
Joplin is the correction to that trajectory. It’s an open-source note-taking app that doesn’t care about “AI-powered insights.”
It uses Markdown, which means your notes are clean and portable.
It offers end-to-end encryption. In an era where every major tech company wants to scrape your notes to train their next model, Joplin is a digital bunker.
And it’s completely free. You can sync it using your own Dropbox or OneDrive, keeping you in control of your data.
Microsoft To Do
If you’ve tried every complex project management tool on the market, from Notion to Monday, you’ll agree they can be a bit of overkill when it comes to the day-to-day chaos of remembering to bring the right gear to baseball practice or keeping a running list of home-office upgrades.
For that, there’s Microsoft To Do.
When Microsoft bought Wunderlist years ago, it eventually managed to port over the best part: the simplicity.
It’s a list. That’s it. No “intelligent sorting” that hides your most important tasks based on an algorithm. And no Copilot integration… yet?
Microsoft To Do is also one of the easiest ways to manage shared lists with a spouse or a team. Whether it’s a grocery run or a quick checklist for a Fast Company draft, it just syncs and pings.
Goodtime
If you’re tired of focus apps that feel more like mobile games—complete with ads, subscriptions, and persistent notifications—you need to switch to Goodtime.
It’s an open-source, minimalist productivity timer that takes the “dumb” philosophy to its logical conclusion with a pure Pomodoro-style timer that’s completely ad-free and tracking-free.
There are no accounts to create, no cloud syncs to manage, and no “intelligent” suggestions to ignore.
Speaking of ignoring stuff: when you start your timer, the app can automatically trigger Do Not Disturb to create a barrier between you and your notifications.
While other apps try to keep you engaged with the screen through gamification, Goodtime encourages you to start the clock, put your phone down, and forget it exists until the work is done.
Composite 25Trend 10
Mythos AI may be a cybersecurity threat, but it follows the rules of the game
The cybersecurity community went on alert when Anthropic announced on April 7, 2026, that its latest and most capable general-purpose large language model, Claude Mythos Preview, had demonstrated remarkable—and unintended—capabilities. The artificial intelligence system was able to find and exploit software vulnerabilities—the most serious type of software bugs—at a rate not seen before.
The news ignited concern among the public, world governments, and the information technology sector about the capabilities of today’s AI to undermine cybersecurity, with some people framing the model as a global cybersecurity threat.
Claiming that it would be too risky to release the model, and that the company had the moral responsibility to disclose these vulnerabilities, Anthropic said it would not immediately offer the model to the public. Instead, it granted exclusive access to tech giants to test the model’s capabilities, a process Anthropic dubbed “Project Glasswing.”
As a cybersecurity researcher, I think Mythos’ capabilities are impressive, but the AI system does not represent a radical departure. Mythos is less a new threat than a mirror reflecting how people behave and how fragile modern systems already are.
What Mythos did
During a controlled evaluation, engineers with minimal security experience prompted Mythos to scan thousands of software codebases for vulnerabilities. The model showed striking capabilities in conducting multistep, autonomous attacks that take experts weeks or even months to put together. Mythos was not only able to discover 271 vulnerabilities in Mozilla’s Firefox, it also developed exploits to take advantage of 181 of those.
Overall, Anthropic’s red team, which takes on the role of an attacker to test defenses, and the United Kingdom’s AI Security Institute reported that Mythos found thousands of zero-day, or previously unreported, vulnerabilities in major operating systems, web browsers, and other applications—software flaws that have not yet been patched and can be turned into exploits immediately. National Security Agency officials testing Mythos have been impressed by the tool’s speed and efficiency in finding software vulnerabilities, according to a news report.
Anthropic’s announcement of Mythos and the cybersecurity threat it poses garnered widespread media attention.
Among the most widely reported were Mythos’ ability to identify a dormant 27-year-old security flaw in OpenBSD, a security-focused operating system, and a 16-year-old bug in FFmpeg, a video/audio processing tool. Some of these flaws allow unauthenticated users to gain control of the machines hosting these applications.
Even more striking, the relatively inexperienced engineers running Mythos’ evaluations were able to use Mythos to complete attacks overnight, from finding vulnerabilities to exploiting them—something that can take human experts weeks to do. The model’s ability to chain multiple steps is what surprised Anthropic and organizations that tried it. In an evaluation by the AI Security Institute, Mythos was able to take over a simulated corporate network in three out of 10 tries, the first AI model to succeed at the task.
These results are real. They also paint an incomplete picture in ways that matter.
Where is the breakthrough?
At first glance, Mythos’ breakthrough sounds novel and could signal a new class of cyber threats. However, a closer look suggests something different. The vulnerabilities Mythos found are not new in nature. They generally don’t belong to unknown security flaws, and in many cases they are variations of well-known and well-understood classes of software vulnerabilities.
In cybersecurity, finding new instances of known types of flaws is not unusual. The most successful attacks rely on known, well-defined vulnerabilities that stay overlooked or unpatched. What concerned the researchers was not Mythos changing the nature of finding and exploiting vulnerabilities, but rather the intense scale and speed with which it was able to find and exploit those vulnerabilities.
This is not a breakthrough per se but rather a result of decades of research in both cybersecurity and AI. In that sense, Mythos is the natural—and expected—result of powerful automation and AI integration because it follows the same fundamental procedures used in standard offensive cybersecurity practices. These include scanning for vulnerabilities, identifying patterns, and testing exploitability. Mythos and similar emerging models make it possible to chain these steps together at a speed that is hard to fathom.
So why were these vulnerabilities missed in the first place?
It is crucial to understand that not all vulnerabilities are cost-effective to fix, and not all vulnerabilities are a priority. Mythos did not discover a new kind of weakness—it exposed the limits of how cybersecurity practitioners search for them.
New tech, age-old dynamic
Mythos highlights an important fact about the reality of cybersecurity threats. System defenders are always at a disadvantage because they need to always succeed. Attackers, however, need to succeed only once to break the security of a system. This cat-and-mouse game will always be the same, and Mythos does not change that—it simply reinforces it.
Mythos follows a familiar dynamic: A tool created to protect can also be used to attack and harm.
“The same improvements that make the model substantially more effective at patching vulnerabilities also make it substantially more effective at exploiting them,” Anthropic officials wrote in a blog post about Mythos.
What once may have required highly specialized skills can now be achieved with significantly less effort, which raises the most important question: Who will benefit first by using tools like Mythos—defenders or attackers?
Mohammad Ahmad is an assistant professor of management information systems at West Virginia University.
This article is republished from The Conversation under a Creative Commons license. Read the original article.
Composite 25Trend 10
AI data center boom squeezes consumer tech’s chip supply—even though they use different chips
The boom in data center construction is taking up much of the supply of high-tech components, especially processor and memory chips. This demand is squeezing consumer device makers, which are having trouble acquiring enough chips.
This is happening even though data center servers and smartphones use different types of chips. The key distinction between consumer electronics and data centers is what they need chips to be optimized for. Smartphones and PCs require low power use, thermal efficiency, and tight integration. Data centers that run AI systems such as large language models, or LLMs, require maximum compute power, memory bandwidth, and storage throughput.
To meet these needs, consumer devices tend to rely on systems-on-a-chip—chips that combine processing and storage—with dynamic random access memory, or DRAM, and NAND, a type of nonvolatile memory. In contrast, AI servers rely on graphics processing units, or GPUs, or other accelerator processors combined with high-bandwidth memory chips.
I study global supply chains and how businesses respond to market constraints within these supply chains. The reason for the consumer electronics supply crunch has to do with the nature of the chip market: its concentration, high costs, and how it responds to boom-and-bust cycles.
AI is not replacing consumer electronics; it is reorganizing the chip market around new priorities for specific chip characteristics. Data centers are pulling capital and scarce memory capacity toward the production of accelerator processors and high-bandwidth memory and the data handling and electronics equipment that surround them.
Chipmaking explained.
A winner-takes-most industry
Chip manufacturing behaves less like a competitive commodity market and more like a layered oligopoly. Scale matters because the leading firms can reinvest in research, improve yields, secure equipment, and deepen customer relationships. In the case of graphics processor chips, designers such as NVIDIA, which has 85% market share, depend on advanced semiconductor foundries such as TSMC, which has more than 70% market share, to manufacture chips using extreme ultraviolet lithography machines from ASML, a monopoly.
A small number of producers both design and manufacture memory chips. Currently, three companies—Samsung, Micron, and SK Hynix—hold a majority market share in the memory chips market. Long development cycles, extremely high fixed costs and the need for technological leadership reinforce concentration over time.
Consumer electronics firms such as Apple, along with other technology firms such as Amazon, Google, Microsoft, and Xiaomi, increasingly design their own processor chips, because these chips shape the user experience, AI performance, power efficiency, and system-level differentiation. Manufacturing memory chips, by contrast, is extraordinarily capital-intensive; requires high precision, efficiency, and production line utilization; and is dominated by a few incumbent suppliers.
Since 2000, the memory chip industry has moved through repeated cycles of overcapacity and undersupply: the post-dot-com collapse, the 2007-09 glut, the tighter 2010s after consolidation, the severe 2022-23 downturn, and the AI-driven tightness of 2024-25. This has led to high levels of concentration in the industry and chipmakers that are hesitant to add capacity. Producers often operate chip fabrication plants, or fabs, at or near capacity due to high fixed costs. The risk of having expensive facilities go underused keeps chipmakers from bringing new fabs online in lockstep with demand increases.
Consolidation has reduced the number of major suppliers, who now increasingly direct investment toward higher-margin products rather than broadly adding capacity. That shift is important for understanding why AI demand is tightening chip supplies even as demand for consumer electronics continues to grow.
The most advanced computer chips are made with a machine manufactured by one Dutch company.
How the AI data center boom redirects capacity
The AI boom has changed memory demand from a broad consumer cycle into a more segmented market centered on high-bandwidth memory chips. In 2023, Micron cut capital spending and the company’s fabs operated below levels needed to justify their cost. By 2026, however, Micron was reporting strong AI demand, record data center DRAM revenue, and rapidly rising high-bandwidth memory sales.
This shift matters because the market for supplying memory cannot respond quickly. Opening new fabs requires years of planning, large capital commitments and investments in advanced process equipment and skills. Memory chip manufacturers are likely to remain cautious about expanding capacity even as their profitability improves, with 2026 spending focused more on technology upgrades and high-value products than on large increases in chip supply.
In practical terms, AI is not simply lifting all memory demand equally; it is redirecting scarce capacity toward massive, or hyperscale, data centers and server markets first.
Can consumer electronics catch up?
Consumer electronics can catch up, assuming the manufacturers can weather the cost increases from tariffs and geopolitical pressures. One way they could is by making investments to enable small AI language models to run on consumer devices, a move analysts expect the companies to attempt.
Apple shifted a growing share of U.S.-bound iPhone production out of China to India and moved much of its iPad, Mac, Apple Watch, and AirPods assembly for the U.S. market to Vietnam to lower the company’s tariff burden. Yet relocation does not eliminate cost pressure. Manufacturing iPhones in India still costs roughly 5% to 8% more than in China, and in some cases closer to 10%, because supplier ecosystems, logistics, and production efficiency remain stronger in China.
Rising geopolitical tensions between the United States and China led to supply constraints and export controls on critical minerals and chip components, raising input costs for consumer electronics manufacturers. This led to higher total import costs and reduced margins for firms unable to pass costs fully to consumers, leading to further consolidation in supply.
Consumer devices do not need to replicate data center infrastructure to offer AI on their products. Their opportunity lies in running small language models on-device for summarization, rewriting, search, assistance, and lightweight reasoning. Doing so, however, creates a distinct hardware requirement. Phones and laptops need to incorporate multiple functions on the same chip, combining processing capability with fast local memory and enough storage to keep on-device AI responsive. Apple’s current device requirements for the company’s AI, Apple Intelligence, also show that older phones often lack the compute power and memory needed for useful on-device AI.
To adopt AI, device makers need to redesign their products with higher-end chips—both processors and memory—that can piggyback on the AI model-oriented growth in the chips market driven by the data center boom. Such a shift by the device makers could also provide a useful backstop for the memory chipmakers in case the projected AI and data center growth does not materialize in the medium to long term, a boom-and-bust cycle that memory chipmakers have had to endure many times in the past.
What this means for the wider economy
The AI and data center boom is redistributing capital, supplier attention and pricing power across the broader economy. Sectors with limited purchasing leverage are especially vulnerable when chip supplies tighten. For example, medical technology accounts for less than 1% of the overall chip market, leaving essential equipment manufacturers exposed during shortages.
In contrast, sectors linked to power delivery and digital infrastructure may benefit from the boom because they try to keep up with demand for cloud services and electrification. The International Energy Agency estimates that data centers consumed about 415 TWh of electricity in 2024 and notes that AI is accelerating the deployment of high-performance servers, which implies stronger demand for the grid, storage, cooling, and networking equipment around them.
For the consumer electronics industry, the strategic task is not to try to match the AI data centers chip for chip but to build differentiated, energy-efficient, on-device AI services while managing higher supply chain and tariff risks.
And for consumers looking to buy phones, games and laptops, because of high demand from data centers, the next few years are likely to bring higher prices, shortages, and delayed product releases.
Vidya Mani is an associate professor of business administration at the University of Virginia and Cornell University.
This article is republished from The Conversation under a Creative Commons license. Read the original article.
Composite 25Trend 10
The American tech manufacturing success story hiding in plain sight
On Wednesday, Nvidia and Corning announced a $500 million deal to build fiber-optic cables to power AI data centers. For Nvidia, which manufactures graphics processing units key to building and training top-tier AI models, the partnership will help the chipmaker reduce latency and energy consumption for AI systems and likely accelerate its move to co-packaged optics. This would have fiber connections more directly integrated with chips.
Per a Securities and Exchange Commission filing, Nvidia now has a pre-funded warrant to purchase 3 million shares in Corning and the option to purchase 15 million more. As part of the agreement, Corning says it will increase its optical connectivity manufacturing tenfold and add more than 3,000 jobs, including at new factories in Texas and North Carolina.
“Their commitment is directly fueling the expansion of our U.S. manufacturing footprint and creating more than 3,000 new high-paying jobs for American workers,” Corning CEO Wendell Weeks said in a statement.
This is all, no doubt, evidence that the AI race is heating up. But it’s also just the latest deal for Corning, a New York-based aspects and materials science company that now plays a critical role in the U.S. technology manufacturing industry. As U.S. officials and tech investors look to pivot to hard tech and bolster the domestic supply chain for advanced manufacturing, it’s notable that Corning has already become integral to this sector. The Nvidia deal is only the latest example.
This is all the more impressive considering that Corning was founded back in 1851 and has remained relevant, even amid remarkable evolution. Its résumé includes designing bulbs for Thomas Edison’s incandescent lamps, introducing the world to Pyrex cooking glassware, and now developing glass used in virtual reality headsets. By modern standards, companies are lucky if they last a few decades. Manufacturing comes with added challenges, including high up-front investments in production lines that can quickly become outdated. This makes Corning a unicorn of sorts.
Consider that earlier this year, the company announced a new deal, worth up to $6 billion, to provide optical cabling and connectivity to Meta, and soon began construction on a new plant in Hickory, North Carolina, that will support its work for the tech company. Corning has also said it has two additional agreements with hyperscale customers that are “similar in size and duration” to the one with Meta, though it hasn’t revealed which ones.
The company has had a spate of deals with a range of other technology companies working to develop next-generation tech. These include agreements with Lumen Technologies (to make optical cables for data centers), Xanadu (a Canadian quantum chip manufacturer), Broadcom (again, to build co-packaged hardware), and solar companies Suniva and Heliene (to make silicon wafers and polysilicon for the only solar panels assembled entirely in the U.S.).
Generally, Corning stands to profit as companies look to phase out copper for fiber.
And then, of course, is the company’s massive business making glass for smartphones and other electronic devices, including its robust Gorilla Glass business. Corning is a major supplier for Apple, and last year the two companies officially agreed to manufacture all iPhone and Apple Watch cover glass in Kentucky. Corning also makes glass for Samsung and Nokia, and has plenty of other business lines, too, including automotive and life sciences work.
Not bad for a 175-year-old company.
Composite 26Trend 10
This Gen Z film distributor is using influencer events to get his peers going to the movies
Peter Gold has always loved making films. While attending film school in New York, he became involved with a film called Our Hero Balthazar, directed by Oscar Boyson, known for his work as an executive producer on Uncut Gems.
Gold instantly knew the film was something special. He also knew it would be tough to find distribution in today’s theatrical marketplace.
The dramedy, starring Jaeden Martell as a wealthy New York City teenager Balthazar Malone, who, eager to impress his activist crush, follows an online connection (Asa Butterfield) to Texas where he believes he can stop an act of violence, was passed over by A24 and Neon.
So Gold, 26, decided to launch his own distribution company, WG pictures, financed through outside investors, with film producer Brad Wyman to make sure it saw theatrical release.
“Filmmaking and storytelling are the heart of my passion. Getting into distribution really came from a place of frustration with the state of independent cinema,” Gold told Fast Company. “So many movies, including my own, were being overlooked by existing distributors and weren’t being given the opportunity they deserved.”
Our Hero Balthazar opened March 27 at Regal Union Square as the number 2 film in the theater, generating $33,138 opening weekend gross, second only to Project Hail Mary. The film’s budget was under $2 million. The film opened sold-out in LA on April 4 and is now expanding across the country.
Hollywood should take note. The amount WG Pictures has spent on distribution is less than $1 million. WG Pictures pulled off the feat without spending a single dollar on paid media and instead relied entirely on social media to drive awareness.
From TikTok fan edits to Letterboxd influencers, social media has proven a boon for cinema. With it, a new kind of showmanship-based marketing has emerged. Cynthia Erivo and Ariana Grande mastered the art of going viral on social media during the Wicked press tour. Timothée Chalamet appeared on a Wheaties box and hosted a table tennis tournament to promote his most recent project, Marty Supreme.
“Honestly, I thought A24 did an interesting job with Marty Supreme, but they have Timothée Chalamet,” said Gold. “We don’t have Timothée Chalamet. We have to work with what we have.”
Gold worked with the filmmakers closely to come up with a social media strategy driven by the characters and the story. They started by creating an Instagram account for the film’s protagonist, with the handle @bboymalone212, that has since amassed more than 72,000 followers.
One post on the Instagram page features a custom starter pack meme inspired by the character of Balthazar, with performative male staples like a New Yorker tote bag, Lorde album cover and wire headphones. Another post features an Erewhon haul of coconut matcha cold foam and Lemme Purr vaginal probiotic gummies, touching on the film’s themes of exhibitionism in the social media age.
“We’re telling the story of this character and building awareness around the movie without just running a trailer with paid ad spend,” said Gold. The social media generation no longer wants to be marketed at, Gold understands, they want to feel like an active participant.
The Instagram account’s most viral post tapped content creator Caleb Simpson, who on his own has more than 2.8 million followers with his viral street series where he asks strangers, and more recently celebrities, “How much do you pay for rent?” and follows it up with, “Can I get a tour of your apartment?”.
Simpson and Martell, in character as Balthazar, joined up for the Instagram Reel, touring the 80th floor New York City apartment overlooking Central Park, which was also a set in the film. “I try not to focus too much on money,” says Martell as Balthazar in the clip. “I’m more focused on making a change.” The comments are a mix of those in on the joke and bemused onlookers, none the wiser. “That was the first time Caleb had ever done a fictional person,” says Gold.
WG Pictures also took advantage of the impressive social media following of those involved in the film, including actress and singer Halsey and actor Noah Centineo, boasting a combined 40 million followers. Each pulled their weight with non-stop posting about the film in the run up to its release, culminating in more than 30 million organic social impressions. Gen Z and Millennials say social media is the number one form of discovery for films, according to a new Fandango study.
Higher ticket prices, the rise of streaming platforms and worsening theater etiquette, have all contributed to deflated box office numbers. A survey from October shows that overall cinema attendance has remained flat since 2019, but the percentage of frequent movie-goers has dropped from 39% to 17% in 2025.
In 2025, 780 million people actually went to the movies according to EntTelligence’s annual report, down from 820 million in 2024. Over the same period, ticket prices jumped 5.7%. Between 2005 to 2019 – before the Pandemic shuttered screens and accelerated a shift towards streaming – the industry averaged well over 1B tickets sold annually.
While Hollywood has expressed its fears that the streaming era and smartphones will stop the social media generation from leaving the house and going to watch films the old fashioned way, in a dark room filled with strangers, the opposite is proving true.
Gen Z is now the most active cinemagoing demographic, according to Fandango, having seen seven films on average in 2025, compared to 5.3 for the general population.
And while millennials mainly treat moviegoing as an escape from daily grind, Gen Z sees it primarily as a social activity. Gen Z also attributes a better selection of movies and the appeal of leaving the home as key motivators for going to the movies.
In the US, 95% of Gens Y and Z are now interested in exploring their online interests through in-person events, according to Eventbrite data. Both Gen Z and Millennials also prefer to extend moviegoing beyond the screen, pairing it with dining and drinking, according to Fandango.
Gold and WG pictures are meeting that audience where they are at. Opening weekend for Our Hero Balthazar, WG pictures hosted a rave at the Museum of Sex in New York City. “I felt like that was something Balthazar would have thrown himself,” says Gold. To gain access, attendees needed a ticket stub for the film. A slightly less extreme marketing stunt than film distributor, Focus Features, who only permitted fans with bald heads (there was a barber in the foyer for those ‘willing to become bald’) for an early screening of sci-fi comedy film Bugonia.
WG Pictures also hosted an immersive gallery experience with visual artist Jet Le Parti, where they created original artwork inspired by the film and the issue of gun violence, reflects WG’s broader strategy of eventizing cinema. They also hosted an event with Third Space–hosted event, designed to convert awareness into active participation and, subsequently, ticket sales.
This social driven strategy is a shift for what has, and still is, a mostly solitary experience. When the lights dim and the film starts rolling, talking or, worse, scrolling, is strictly forbidden. And yet, Gold is banking on community being the next big drive getting Gen Z to the box office. “It’s not just cinema in a crowded theater,” as Gold sees it. “It’s an opportunity to connect with the community.”
The success of platforms like Letterboxd and WG picture’s IRL marketing strategy is a testament to that. “Someone said to me after one of the screenings at Roxy Cinema that this is a movie that starts after it’s over,” he explains. “In terms of the conversation it provokes.”
For Gold, the biggest challenge isn’t getting Gen Z to the cinema, it’s finding the right movies. “We’re working on Toad, which is a stoner comedy, and looking at some really interesting documentaries,” he said of future releases. “But it’s really just about finding the next exciting movie and continuing to distribute films theatrically.”
Find the right movie, market it right, and Gen Z will come.
Composite 26Trend 10
A PC trade-in rush is on the way—and it’s coming at the worst possible time
Just as they did with televisions, many people used the pandemic as an excuse to upgrade their PC or laptop.
It was a move that made sense at the time. Telecommuting became essential, and not all devices could adequately handle the demands of Zoom, Teams, and other work software. At the same time, digital communication was often the only way to stay in touch with friends and family.
Smartphones handled some of that heavy lifting, of course, but the PC industry still saw shipments spike 14.5% from 1999 to 2000.
Now, much like the TV market, many PC owners are reaching the point where a new device is becoming necessary. But unlike that living room fixture, PC shoppers are entering a hostile market defined by higher prices and fewer meaningful performance gains.
IT research firm Gartner notes that many people replace their business devices, typically laptops, every three to five years. International Data Corp. puts that timeline closer to five to eight years when businesses actively manage upgrades and repairs. Personal-use computer owners tend to follow a similar replacement cycle.
That means a refresh wave is looming for pandemic-era buyers, just as component prices are soaring amid AI-driven demand for hardware. RAM prices have jumped anywhere from 150% to more than 200% over the past year, depending on the type, according to PCPartPicker.com. Storage prices, including the cost of hard drives, have followed similar trends.
Meanwhile, video card prices have remained elevated for years, as GPUs, the chips that power graphics cards, have become a core component of AI systems. For gamers, that has been especially frustrating.
PC gaming is rapidly becoming a more important part of the video game ecosystem, threatening to displace consoles, according to some industry leaders at the recent Iicon conference hosted by the Entertainment Software Association. Analysts, however, say Nvidia is not expected to release a new generation of its GeForce GPUs in 2026. If that happens, it will mark the first time in three decades the company has skipped an annual release cycle. And finding a top-of-the-line RTX 50-series card remains difficult for many enthusiasts, with some retailers charging double the suggested retail price.
A vanishing entry level
As frustrating as the price hikes already are for consumers in need of an upgrade, analysts do not expect the situation to improve anytime soon. A separate Gartner projection predicts that PC prices will rise 17% this year compared with 2025. Worse still for consumers simply looking for a functional home computer, the era of low-cost machines may be nearing its end.
“The sub-$500 entry-level PC segment will disappear by 2028,” says Ranjit Atwal, senior director analyst at Gartner. “In addition, rising AI PC prices will delay the projected 50% market penetration of AI PCs until 2028.”
PC vendors, Gartner says, are likely to accept lower sales volumes to protect profit margins rather than aggressively pursue price-sensitive customers, noting that the first half of this year represents a “critical window.” By the end of the year, the firm predicted, combined prices for DRAM and solid-state drives could rise 130%.
The surge in component costs, combined with uncertainty over how long those increases will last, could reshape the U.S. computer refresh cycle in one of two ways.
Some analysts believe laptop users may simply hold onto devices as long as they remain “good enough” to run everyday programs and apps. Desktop users with some technical know-how can also upgrade individual components at a lower cost, or turn to services like Geek Squad if opening up a PC feels too intimidating.
Others argue that buyers may rush to upgrade now before prices climb even higher. Distributors appear to be betting on that scenario. Worldwide PC shipments rose 4% in the first quarter of 2026, to 62.8 million units. That increase is notable because 2025 figures were already inflated as companies front-loaded inventory ahead of the Trump tariffs.
“The 4% year-over-year PC shipment growth in the first quarter of 2026 was artificially inflated,” says Rishi Padhi, research principal at Gartner, in a statement. “It was not due to genuine demand, but instead because of vendors’ and channel distributors’ increase of inventory levels ahead of expected price hikes in the second quarter.”
Composite 27Trend 10
Grok’s usage is so low that Elon Musk can sell compute to Anthropic
Welcome to AI Decoded, Fast Company’s weekly newsletter that breaks down the most important news in the world of AI. I’m Mark Sullivan, a senior writer at Fast Company, covering emerging tech, AI, and tech policy.
This week, I’m focusing on Elon Musk’s decision to lease the computing capacity at SpaceX’s Colossus 1 data center to Anthropic. I also look at what a new Atlantic exposé on David Sacks says about Silicon Valley’s alliance with Trump, and a benchmark that’s stumping top AI coding agents.
Sign up to receive this newsletter every week via email here. And if you have comments on this issue and/or ideas for future ones, drop me a line at [email protected], and follow me on X (formerly Twitter) @thesullivan.
Why Grok is selling compute to Anthropic
While everybody else in the AI space scrambles to lock down computing power, xAI’s Grok models are apparently being used so little relative to peers that the company can sell off the capacity of entire data centers, “colossal” ones at that.
Anthropic said Tuesday it had signed an agreement with SpaceX to use all the computing capacity in SpaceX’s Colossus 1 data center in Memphis. (SpaceX owns xAI.) The deal will give Anthropic access to more than 300 megawatts of computing capacity, or more than 220,000 NVIDIA GPUs. Anthropic says the additional capacity will be used to serve its Claude Pro ($20 per month) and Claude Max ($100 to $200 per month) subscribers.
SpaceX CEO Elon Musk says he gave his much-sought moral stamp of approval to Anthropic. “By way of background for those who care, I spent a lot of time last week with senior members of the Anthropic team to understand what they do to ensure Claude is good for humanity and was impressed,” Musk said in an X post. “Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector.”
Musk says xAI had already shifted its training workloads to Colossus 2, freeing up Colossus 1 for Anthropic’s use. Anthropic says it will use the facility primarily for inference, or the processing required to respond to user prompts in real time.
The partnership could eventually extend beyond Earth. Anthropic says it has also been discussing plans with Musk and SpaceX to develop multiple gigawatts of orbital AI compute capacity. Space-based AI data centers hold obvious appeal because the cost of cooling servers would essentially disappear. But major technical hurdles remain, especially around reliably transmitting massive amounts of data between orbiting infrastructure and Earth.
Musk’s willingness to arm Anthropic with vital computing power may also have something to do with his hatred of Anthropic rival OpenAI, and his dislike of OpenAI founder Sam Altman. Musk sued OpenAI, claiming the company’s leadership betrayed its original nonprofit mission to develop AGI for the benefit of humanity rather than for profit.
Trump’s bargain with Silicon Valley on AI may be weakening
The Atlantic’s George Packer, in a new article about former White House “crypto and AI czar” David Sacks, sheds more light on how and why Sacks and other Valley elites went full MAGA before the 2024 election. Now there are signs that the main thing Silicon Valley wanted in exchange for its support may be in jeopardy.
Silicon Valley’s preferred version of its MAGA conversion story is that influential VC Marc Andreessen met with representatives of the Biden administration and was told the administration intended to heavily regulate AI so that only a few big AI labs, and no startups, would be able to comply and survive. Andreessen said Biden wanted to “nationalize or destroy” Silicon Valley. He said Biden wanted to kill the entire cryptocurrency industry. He said he and his partner Ben Horowitz decided to support MAGA right after that meeting.
Biden officials dispute Andreessen’s account of what was said. But Andreessen’s version was enough to set a broader shift in motion among tech elites. Sacks held a fundraiser for Donald Trump in June 2024 in San Francisco’s wealthy Pacific Heights neighborhood. After talking with Trump at the event and on the All-In podcast, Sacks said: “All of his instincts are Let’s empower the private sector; let’s cut regulations; let’s make taxes reasonable; let’s get the smartest people in the country; let’s have peace deals; let’s have growth.”
What Sacks and others were really after was a promise of AI deregulation and more tax cuts. They got the tax cuts, and so far the Trump administration has worked hard to stifle government investigations or regulations targeting the tech industry. Some states have passed laws requiring government oversight, but the administration has been trying to preempt such laws or challenge them in court.
Packer suggests that Sacks, Andreessen, Horowitz, and other Valley elites may also share something in common with much of MAGA: They are white men witnessing a loss of status in society. “Andreessen was willing to pay high taxes and support liberal causes and candidates as long as he was regarded as a hero,” Packer writes.
But Silicon Valley’s fall from grace is not the fault of Democrats, Biden, or “wokism”; it’s the result of government and society slowly realizing that many Silicon Valley elites are not actually driven by idealistic notions of “making the world better.” Instead, they’ve repeatedly shown a willingness to unleash technologies they know may be harmful. The clearest example is Meta, which the government largely allowed to regulate itself while shielding it from many user lawsuits through Section 230, only to watch social media platforms contribute to disinformation, political polarization, and harms to children.
But nothing is permanent with Trump, as so many others have found out, and agreements that no longer provide immediate value can be quickly abandoned.
The White House announced this week that it’s considering a requirement that government officials “vet” new AI models before they can be released. Team Trump was apparently spooked by two things. An AI model from a company it recently declared a supply-chain risk, Anthropic, developed a model called Mythos that can identify software vulnerabilities at scale and devise ways to exploit them. Meanwhile, backlash against the tech industry’s massive data center buildout is becoming increasingly unpopular with parts of the MAGA base and could become a major GOP liability in the midterms.
Maybe tech elites and MAGA don’t mix quite as well as either side once thought.
Meet the new benchmark that’s soundly defeating coding agents
Perhaps the most consequential application of generative AI models so far has been software engineering, where agents generate code and increasingly make high-level architectural decisions. But how do we tell how good an AI software engineer really is? Until now, the industry has largely relied on benchmark tests such as SWE-Bench, which evaluate models on relatively well-defined tasks like fixing bugs or implementing a single feature. Now the developers behind SWE-Bench have introduced a much harder test called ProgramBench.
The benchmark is difficult because the AI agent has to reason strategically about the optimal architecture and programming language needed to reproduce the performance of each of the 200 test programs. Once an agent finishes building a codebase, the benchmark runs roughly 248,000 tests to measure how closely the recreated software matches the original behavior.
So far, all of the major models tested on ProgramBench, including Anthropic’s Claude Opus 4.7, Google’s Gemini 3 Pro, and OpenAI’s GPT-5.4, have scored big fat zeros. In other words, none were able to fully complete the test builds. Several models, however, were able to complete portions of them.
The results suggest that current AI coding tools still are not advanced enough to make the kinds of architectural and systems-level decisions human software engineers routinely make when turning an idea into working software. The findings may also indicate that AI agents still struggle to apply abstract principles learned during training to entirely novel problems.
More AI coverage from Fast Company:
How a Texas vegan cheese-maker used Claude and Manus to fight back against a big shipping company
AI power users are pulling away from everyone else, Microsoft says
AI labels were supposed to help users spot fakes. Here’s why they’re failing
OpenAI’s trillion-dollar AI bet is a study in ‘riskmaxxing’
Want exclusive reporting and trend analysis on technology, business innovation, future of work, and design? Sign up for Fast Company Premium.
Composite 49Trend 10
5 free, pro-level PC and Mac apps to replace your paid subscriptions
If you’re like me, your bank statement looks like a graveyard of monthly $9.99 charges for apps and web services that somehow add up to the price of a used Honda Civic every year.
Somewhere over the last decade or two, software companies turned us from owners into renters. And quite frankly, the landlords are getting greedy.
But here’s the good news: Whether you’re on a Mac or a PC, there are world-class alternatives that don’t require a monthly tribute to a corporate overlord.
We’re talking professional-grade tools that are either free forever or have free tiers so robust you’ll forget the paid version even exists.
Stop renting your digital life. Here are five free, fully-functional apps to check out.
Affinity for design and publishing
For years, the advice was: “If you want to do real design, pay the Adobe tax for Photoshop, Illustrator, and InDesign.” Not anymore.
Since Canva took the reins at venerable graphics app maker Affinity, it’s done the unthinkable and made the core Affinity suite of Photo, Designer, and Publisher completely free.
This isn’t some “lite” version that watermarks your exports. It’s the full, professional experience.
You’re getting layers, masks, vector tools, and desktop publishing without the $600-a-year Creative Cloud bill.
If you need the fancy cloud-based AI tools, sure, there’s a paid option. But for 95% of us, the free version is a total no-brainer.
LibreOffice for word processing, spreadsheets, and presentations
Microsoft 365 wants your money every single year just so you can write a memo or balance a spreadsheet. LibreOffice is the open-source hero that keeps that money right in your pocket where it belongs.
It handles Word docs, Excel sheets, and PowerPoint decks with ease, and it doesn’t need an internet connection or a Microsoft account to function.
LibreOffice might not be as shiny as the latest web-based office suites, but it’s fast, stable, and respects your privacy. It runs on Windows, macOS, and Linux, so you’re covered regardless of your platform.
DaVinci Resolve for video production
If you’re still paying for expensive video production suites, you’re doing it wrong.
DaVinci Resolve is the same software used to color-grade actual Hollywood blockbusters, and the free version is mind-bogglingly capable.
You get professional editing, advanced color correction, and Fairlight audio tools for the low, low price of zero dollars.
Unless you’re exporting in 8K resolution or need specific high-end grain filters, the free tier will do everything you need and more. It’s arguably the best free deal in the entire tech industry, period.
Raycast for productivity tools
Mac users used to swear by Alfred, but the best features were locked behind a paywall. Raycast changed that, and now the Windows crowd finally gets to join the party with the new Windows release.
It’s a productivity monster that replaces your calculator, your window manager, your clipboard history tool, and about a dozen other single-purpose apps.
Raycast is faster than macOS’s Spotlight or the standard Windows Start search, and infinitely more customizable. You can uninstall half of your utility apps once you get the hang of the command bar. It’s the single best thing you can do for your workflow without reaching for your wallet.
Obsidian for notes and organization
Stop paying for Evernote only to have your notes held hostage behind a login screen or a “storage limit.” Obsidian is a markdown-based note-taking app that stores everything locally on your computer.
The free version is essentially the full version: there’s no limit on how many notes you can have or how many plugins you can install.
It’s fast, it’s private, and it allows you to build a massive second brain of interconnected ideas. It’s the ultimate tool for writers and researchers who want to own their data.
