Skip to main content

Fast Company Technology

‘They don’t understand hackers’: Cybersecurity pros explain why everyone panicking about AI-powered hacks is wrong

6 days 18 hours ago

As head of counter-adversary operations at CrowdStrike, Adam Meyers has a front-row seat to the threats facing companies. Yet even for him, a recent trip to South Korea was out of the ordinary. As he was preparing to sleep on the flight from the United States, Meyers received a message from colleagues informing him that six Korean banks had been hacked and that he would likely face questions about the incident upon landing.

By the time his flight touched down, his team at CrowdStrike had pinpointed the server used by the attacker, recovered records of their activity, and even identified a suspect. The alleged hacker had inadvertently left personal information behind while using AI tools to conduct the attacks.

The incident, which CrowdStrike subsequently investigated, involved Artex, an open-source AI-powered penetration-testing tool, alongside LLMs from a variety of providers. The scale and speed of the attack illustrate why artificial intelligence has fueled growing fears about the future of cybersecurity.

But Meyers isn’t panicking. “I thrive on chaos,” he says. And in a conversation with Fast Company at the company’s Fal.Con conference in Barcelona, he argues that much of the alarm surrounding AI-powered hacking is misplaced. “Fundamentally, this is nothing new. It’s just speed and volume,” he says. “AI hasn’t come up with new exploit paths or new vulnerability classes, right? It’s all the same stuff we’re already aware of.”

That doesn’t mean the technology is harmless. But Meyers believes the cybersecurity industry risks confusing the capabilities of increasingly sophisticated AI models with the actual threat they pose to organizations.

He’s not alone. Ciaran Martin, the former head of the U.K.’s National Cyber Security Centre, argues that much of the discussion around AI-enabled hacking has become detached from the realities of cybercrime. “To some extent, we’ve lost our mind in 2026 because of the extent of the capabilities,” he says.

Martin believes companies developing frontier AI models have focused heavily on what their systems can theoretically accomplish without paying enough attention to the motivations and constraints of actual attackers. “They don’t understand threat actors. They don’t understand hackers. They don’t understand the economics and the sociology and the politics of hacking,” he says.

The result, Martin suggests, is an industry increasingly alarmed by suspicious activity, even as the growing volume of automated attacks doesn’t necessarily translate into more successful compromises. That disconnect is putting pressure on security teams, which must sift through an ever-growing number of potential threats.

Eerke Boiten, a cybersecurity professor at De Montfort University in the U.K., sees a similar challenge. AI is accelerating the discovery of vulnerabilities, he explains, but the ability to exploit them isn’t necessarily advancing at the same pace. That leaves already overwhelmed security teams struggling to determine which threats warrant their attention. There is, however, a potential upside: The same AI capabilities that help attackers identify vulnerabilities could also help defenders fix them and assess how difficult they are to exploit.

When every vulnerability becomes an emergency

For Meyers, the challenge underscores a long-standing problem with how organizations assess cybersecurity risks. He argues that the industry has become too reliant on the Common Vulnerability Scoring System, or CVSS, which assigns numerical scores to security flaws on a scale of one to 10. PHP2Shell, a pair of vulnerabilities in the widely used PHP programming language that drew widespread attention, each received a score of 9.2.

Meyers says CrowdStrike examined the issue before its public disclosure and concluded that exploiting the vulnerabilities would be far from straightforward. University of Surrey cybersecurity professor Alan Woodward agrees, noting that doing so required exceptionally precise timing and chaining vulnerabilities together. Still, the flaws weren’t merely theoretical: The Cybersecurity and Infrastructure Security Agency, part of the U.S. Department of Homeland Security, added both to its catalog of known exploited vulnerabilities on October 26, calling for urgent action.

Woodward maintains that CVSS remains useful because it accounts for the potential impact of an attack. Meyers, however, argues that severity scores shouldn’t be the sole factor determining how defenders respond. Instead, organizations should also consider evidence of attacks taking place in the wild. “I like to think about things in terms of what’s being actively exploited,” he says.

Otherwise, the industry risks creating a boy-who-cried-wolf scenario. “The more people get burned out, the more they’re going to make mistakes,” Meyers warns.

Boiten, for his part, sees reason for optimism. He believes that the collaboration underpinning vulnerability identification could eventually extend to fixing flaws, with AI-assisted coding potentially reducing the number introduced in the first place. That’s a notable shift in his thinking. Just a year ago, he expected the rise of vibe coding to produce even more vulnerabilities.

Amid the growing focus on AI, however, it’s easy to overlook one of cybersecurity’s most persistent vulnerabilities: humans. For all the technological advances transforming cybercrime, attackers still frequently gain access by exploiting human weaknesses.

Who benefits from the AI panic?

So why has AI captured so much attention? Meyers believes the companies developing the technology have played a significant role. “There’s a lot of fearmongering going on, and I think the frontier labs have been stoking that fire for their own benefit,” he says.

Meyers suggests that frightening predictions about AI capabilities could serve the commercial and regulatory interests of frontier AI companies, particularly as they consider going public. By emphasizing the potentially catastrophic consequences of their own technology, these companies may also be reinforcing perceptions of their power and importance.

The danger isn’t simply that organizations might devote too many resources to the wrong threats. It’s that a steady drumbeat of warnings could make them less responsive when a truly serious threat emerges.

“I think we need to be careful not to overhype it or we’ll find people stop listening, and then there will be something they really need to take care of,” says the University of Surrey’s Woodward.

Yes, government officials confirm, the Jensen Huang scandal is really, really bad

1 week ago

Nvidia CEO Jensen Huang has found himself embroiled in a scandal involving an alleged affair, a suspected Chinese spy, and some of America’s most closely guarded technology. Not exactly your garden-variety corporate indiscretion.

There’s a lot to worry about. Experts are concerned that Huang’s repeated calls to loosen restrictions on advanced chip exports to China may have been influenced by the relationship. The affair also raises questions about Nvidia’s dealings with Megaspeed, a Singapore-based customer led by Alice Huang. The Commerce Department began investigating Megaspeed last year, The New York Times reported, but transferred the probe to the Justice Department after Nvidia reached out. (Though they share a surname, Jensen Huang and Alice Huang are not related.)

One of the stranger twists in the still-unfolding scandal is that the Trump administration appears to have known about the relationship—and Alice Huang’s suspected ties to the Chinese government—for some time. Yet intelligence officials apparently never warned Jensen Huang, and Trump seems to be brushing off the scandal. Despite an FBI investigation, the president has called Nvidia’s CEO an “honorable guy” and says he believes Huang didn’t know Alice was a spy. Nvidia’s board, meanwhile, says it has “no reason to believe that either national security or Nvidia secrets were compromised” and has “confidence in Jensen and his leadership team.”

But current and former government tech officials say the assurances from Trump and Nvidia’s board do little to dispel their concerns. “A sophisticated individual should not have fallen prey to these advances,” a senior government official who focuses on AI issues tells Fast Company.

‘Obvious targets for espionage’

The concern extends beyond Jensen Huang’s call for U.S. regulators to clear advanced chips to China. It’s possible Alice Huang may have picked up classified information—anything Trump might have mentioned—from the Nvidia CEO, given his close relationship with the president, suggests David Nesting, a former White House AI and cybersecurity official. Alice Huang may have had access to Nvidia information, too. (CHIPS for America, the Commerce Department initiative overseeing federal investments in semiconductor manufacturing and research, did not respond to Fast Company’s request for comment.)

“His home, office, electronic devices, and private and professional communications would all be attractive intelligence targets,” Nesting tells Fast Company. “China would be enormously interested in Nvidia’s intellectual property, technology, road maps, supply chain, and how they might circumvent their export controls.”

People who have sensitive and classified information are supposed to protect that information, but also work with security officials and alert them to anyone who might be asking about it, the senior government official says. “This is not just personal foibles,” the official adds. “Nvidia’s power and integral role in AI means that all decisions made by Mr. Huang inure [accrue] not only to Nvidia but to the AI industry.”

The affair and its corresponding national security implications underscore the risks of concentrating so much of the AI industry in the hands of a single company. Nvidia controls an estimated 80% of the market for graphics processing units (GPUs) used in artificial intelligence, according to the Organisation for Economic Co-operation and Development (OECD), an intergovernmental organization with 38 member countries. That dependence has, in effect, exposed the U.S. to an “enormous single point of failure,” argues Varoon Mathur, who formerly worked on AI issues at the White House.

“When so much depends on one company, that company and its leadership become obvious targets for espionage, coercion, and foreign interference,” he says.

‘The wake-up call we all need’

Pamela Jennings, a former senior policy officer at the CHIPS Research and Development program housed inside the National Institute of Standards and Technology, sees the immediate fallout as a corporate crisis for Nvidia’s board to address. But the scandal also raises broader questions about how the U.S. cultivates a more resilient tech sector.

“It’s kind of interesting to think about where we are in terms of—particularly in the U.S.—looking at innovation and how we are supporting and resourcing innovation,” Jennings says. “It’s not so much whether Nvidia has a lot of power or not a lot of power, but what feeds into those larger companies and branches from them to make us even more robust and more of a leader globally.”

For another former AI official at the Transportation Department, the scandal raises an even more fundamental concern: how much influence unelected technology executives wield over American foreign policy.

“We’ve handed the keys of international statecraft to CEOs who answer to no voter,” the official says. “The fact that one executive’s bedroom partner could impact worldwide AI and data sovereignty initiatives is the wake-up call we all need.”

DOJ sues Lyft for allegedly denying rides to people with disabilities

1 week ago

The Justice Department has filed a lawsuit against Lyft, accusing the company of discriminating against passengers with disabilities who seek rides on its platform.

The federal lawsuit, filed Monday, says Lyft “is aware of, but fails to prevent and remedy” several discriminatory practices against riders who need accommodations under the Americans With Disabilities Act. That includes denying rides to people who have service animals or use stowable wheelchairs and refusing to let passengers with mobility issues sit in the front seat, the suit alleges.

“Individuals with disabilities should be able to rely on ride-share services free of discrimination,” said Jesus A. Osete, acting assistant attorney general, in a statement. “We intend to uphold the ADA’s promise that people with disabilities are entitled to equal access in every part of public life, including essential transportation services.”

Lyft firmly rejected the accusations in a statement.

The lawsuit is the latest from the federal government that targets ride-hailing companies for the treatment of would-be riders with disabilities. The administration filed a similar lawsuit against Uber last year, which is ongoing. Uber settled a separate suit in 2022, and Lyft settled a different one in 2020. The companies stress that their policies and in-app tools require that drivers provide ADA accommodations and remind them of the consequences of refusing to do so.

The new lawsuit against Lyft named several examples of people across the country who hailed a ride on the platform but had one—or, in some cases, several—drivers refuse to take them to their destination, including as recently as last month. Those individuals experienced delays in getting to medical appointments, cancellation fees, and other issues as a result, the suit said.

Lyft policies require drivers to accommodate riders who travel with service animals or mobility devices, and “drivers who violate these policies face serious consequences, including permanent deactivation,” the company said in a statement.

“We’ve cooperated with the DOJ throughout its review and remain committed to serving every rider, including riders with disabilities,” the company said.

The lawsuit was filed in U.S. District Court for the Northern District of California. The Department of Justice said it is seeking $53 million for individuals who have been discriminated against and have submitted complaints to Lyft or the department.

—By Hannah Fingerhut, Associated Press

The best strategy to protect against getting hacked is surprisingly old-school

1 week ago

The recent reports of artificial intelligence agents gone rogue—hacking into corporate systems without being told to by the humans who manage them—are enough to make any systems administrator’s hair stand on end.

The increasingly powerful abilities of frontier and other advanced AI models, combined with the apparent inabilities of their creators to keep them caged, have only added to the worries of companies already wrestling with how to protect themselves against AI-armed attackers.

Cybercriminals of all kinds are using AI tools in much the same way legitimate organizations do, boosting their productivity and moving their operations from human to machine speed.

Meanwhile, businesses have understandably embraced AI themselves. But every new AI tool or application a company chooses to roll out comes with its own security risks. 

When it comes to AI, many companies feel like they have to take an “adopt or die” business perspective, says Leslie Nielsen, executive vice president and chief information security officer at Mimecast. “They think everybody’s going to pass them by,” Nielsen says. “But then the other side is, if you adopt too fast, it’s adopt and die, right?”

While AI threats may be rapidly advancing, so are AI-powered defenses. And experts say following old-school best practices can go a long way toward protecting against even the most advanced attackers.

Make sure you’ve got the basics covered

The tactics used by AI attackers aren’t a whole lot different from those of human-powered adversaries, they’re just faster and more efficient, Nielsen says. As a result, the fundamentals of defense don’t need to be much different either.

“It’s all the basic stuff you should be doing anyway,” he says. “It just becomes more important because the attacks are faster and the attacks are cleaner and more successful sometimes.”

That means making sure that access to sensitive data is limited to those who actually need it; immediately patching security vulnerabilities when they become known; and making sure that all critical data is backed up and accessible if needed. And of course, follow the absolute basics, like making sure your cybersecurity defenses require strong passwords and multifactor authentication, and that you’re screening for phishing emails.

Just like the attackers, defenders need to bolster themselves with AI tools to speed up their responses, Nielsen says. He added that this is especially true for small- and medium-size businesses that are probably working with smaller security teams.

Complicating this already hefty task is the ability of advanced AI models to chain together low-level vulnerabilities into previously undiscovered exploits that could be used to breach a system, says Simon Jelley, vice president of product management at Dell Technologies. 

“But those risks have always been there,” Jelley says. “It’s just accelerating those exposures and also making AI ultimately smarter.”

Lock down your own AI

The AI tools used by companies can end up being the insider threats that no one thinks about, experts say. In the case of the AI models gone rogue, it appears they weren’t trying to do anything malicious, just accomplish the objectives set out by their human creators at all costs.

“The difference between AI and a person is AI is the eager intern that’s going to do what you ask, but it’s going to keep working 24-seven to get it done, and it’s going to do everything it needs to do,” Nielsen says.

That could include doing things like digging up login credentials and accessing data it shouldn’t, potentially putting that data at risk of exposure.

That’s why it’s important that there are guardrails in place to keep AI agents on task and to limit their access to sensitive data. But sometimes companies prioritize productivity over safety, Darktrace president and CEO Ed Jennings says.

“You want AI to do all those things that you want employees to do,” Jennings says. “The problem is, a lot of times those agents are given unfettered access. You have to start to watch their behavior.”

Companies should also create acceptable use policies for their AI tools, just like they do for other kinds of technology. Jennings notes that employee use of AI agents for personal reasons isn’t just a security risk, it can also get expensive if humans start burning through tokens. 

Accept that you will be breached

While traditional perimeter defenses can help keep attackers out, they don’t do anything to spot or stop them if they do get in. And given the increasing volume and sophistication of AI-powered attacks, experts say it’s more of a question of when, rather than if, companies will be breached.

Jennings says that’s one of the reasons his company has focused on behavioral detection for more than a decade. Darktrace’s products were originally designed to spot attackers inside systems by looking for anomalies in human user behavior. Now they’re being used to flag the same kind of signs in AI activity.

“AI doesn’t think in terms of boundaries,” Jennings says. “These agents don’t stop at the email layer. They don’t stop at the network layer. They just move laterally. They move freely. They do whatever is necessary and figure out how to do that very quickly.”

Jelley, of Dell, says companies have come to realize that they can’t just say that they want to protect everything, because that’s not realistic. Instead, they’re better off identifying their core operations and data and creating a plan for getting them back up and running should an incident occur.

“Ultimately, it’s about whether you have a mitigation plan in place to fall back on, a resilient infrastructure that you put in place to recover should the attack happen,” he says.

—Bree Fowler

This article originally appeared on Fast Company’s sister website, Inc.com. 

Inc. is the voice of the American entrepreneur. We inspire, inform, and document the most fascinating people in business: the risk-takers, the innovators, and the ultra-driven go-getters that represent the most dynamic force in the American economy.

Blockbuster video game ‘Grand Theft Auto VI’ is almost here. What makes this latest chapter different

1 week ago

Grand Theft Auto has never been much for following the rules.

Nearly three decades after a video game built around stealing cars, outrunning police and embracing criminal mayhem first arrived, the franchise has survived controversy, transformed open-world gaming and become one of the biggest sellers in entertainment.

Now, after a 13-year wait since its last main installment, Grand Theft Auto VI is preparing for another big score. The latest chapter arrives Nov. 19 with a return to the neon-soaked streets of Vice City with two protagonists.

For longtime fans, GTA VI is a highly anticipated return. For everyone else, here’s what to know about the blockbuster video game, its characters and the enormous business riding on its release:

Why is Grand Theft Auto such a big deal?

The franchise began in 1997 as a relatively simple, top-down game that allowed players to roam fictional cities, steal vehicles, evade police and commit crimes.

The release of Grand Theft Auto III in 2001 helped redefine open-world gaming. It gave players greater freedom to explore a sprawling three-dimensional city, pursue missions and interact with their surroundings.

Later installments expanded that freedom with larger environments, cinematic storytelling, social satire and detailed worlds. The influence reached far beyond games about stealing cars.

Ubisoft’s Assassin’s Creed series brought open-world exploration to historical settings, while Bethesda’s The Elder Scrolls games expanded their own longstanding tradition of open-ended fantasy role-playing. Nintendo’s The Legend of Zelda: Breath of the Wild later emphasized experimentation and discovery, allowing players to use the environment itself to solve problems.

For Grand Theft Auto: San Andreas and Grand Theft Auto V, Rockstar Games recruited real-life gang members to help bring authenticity to background characters and dialogue. Former writer and audio director Lazlow Jones has said some performers challenged scripted lines, prompting developers to let them use language drawn from their own experiences.

That attention to detail helped create the sense of a living world that became a hallmark of the franchise, even as the series drew criticism for its depictions of violence, crime and women.

But controversy did not stop the franchise’s commercial rise.

According to publisher Take-Two Interactive, the franchise has sold more than 470 million units worldwide. Grand Theft Auto V, released in 2013, accounts for more than 230 million of those units and reached $1 billion in retail sales in just three days.

Its online multiplayer component, Grand Theft Auto Online, has helped keep players engaged through years of updates, new activities and in-game purchases.

What is the story behind Grand Theft Auto VI?

This time, the action returns to Vice City, Rockstar’s fictional version of Miami, and expands across the surrounding state of Leonida, a Florida-inspired setting filled with beaches, highways, nightlife and criminal enterprises.

At the center are Lucia Caminos and Jason Duval, a couple trying to improve their circumstances through a life of crime. Lucia has recently been released from prison while Jason, a former soldier who became involved with drug runners in the Leonida Keys, is looking for a different path.

Their plans become more complicated when a seemingly straightforward criminal job goes wrong, drawing them into a conspiracy that stretches across Leonida.

The trailers suggest a story mixing romance, robberies, betrayals and the series’ familiar satirical take on American culture.

Rockstar has revealed several supporting characters, including bank robber Raul Bautista, aspiring music executive Dre’Quan Priest and nightclub owner Boobie Ike.

Who is in the cast?

Rockstar has kept much of the voice cast under wraps, including the performers behind Lucia and Jason.

One familiar actor has confirmed his involvement: Stephen Root, known for his work in “Office Space,” “King of the Hill” and “Barry.” He told The Associated Press in September that he appears in the game, adding that fans had already recognized his distinctive voice in promotional footage.

Root has been identified in reports as the voice of Brian Heder, an aging drug runner operating out of the Leonida Keys. Rockstar has introduced the character but has not publicly confirmed Root’s casting in that specific role.

What makes this installment different?

Beyond the return to Vice City, one of the biggest changes is the focus on two protagonists whose relationship drives the story.

Grand Theft Auto V featured three playable main characters. The new installment centers on Lucia and Jason as partners navigating the criminal underworld together.

Rockstar has promoted a larger, more detailed setting that stretches beyond the city into the wider state of Leonida. The game also arrives in a vastly different technological landscape from its predecessor, with newer consoles capable of supporting more elaborate environments and character interactions.

But not every detail is known. Rockstar has offered trailers and promotional previews, while questions remain about the full range of activities, how the game’s systems will work and the future of its online multiplayer experience.

What about the music?

Music has long been part of Grand Theft Auto’s identity. It has included everything from fictional radio stations to the songs accompanying players as they cruise through the game’s cities.

Grand Theft Auto VI is expanding that tradition with an original score by Justice, the Grammy-winning French electronic duo of Gaspard Augé and Xavier de Rosnay.

Rockstar says the score adapts to players’ actions and experiences throughout the game, shifting with the environments and situations they encounter.

A 30-track score album is scheduled for Dec. 11, featuring collaborations with artists including Tame Impala, SZA, King Gizzard & The Lizard Wizard and Gesaffelstein.

The game also features in-game radio stations and a separate collection of original songs, continuing the franchise’s connection to popular music.

—Jonathan Landrum Jr., AP Entertainment Writer

The weird and brilliant innovations that led to electroconvulsive therapy—and the mystery of why it works

1 week ago

When I began considering electroconvulsive therapy, better known as electroshock, I assumed the “shock” part referred to the electrical zap you received. An electrode or two on your head. Some doctor pressing a button. A machine buzzed audibly. Zap, zap. You were shocked. Maybe changed forever. Maybe for good, maybe for ill. Zap, zap. Like a ray gun in a dystopian thriller. And no thriller was more dystopian (I flattered myself) than the anxious depression that had me in agony every waking moment.

Before I met those electrodes, accepted them onto my temples and into my life, my doctors gave me some books to read. I learned that most of what I thought I knew about ECT was deeply wrong. The real story was both grislier and more hopeful than you probably know. 

THE FIRST SHOCK TREATMENTS: FEVERS, INSULIN, AND ELECTROCUTED PIGS

When the Romanian-born psychiatrist Constance Pascal began using the term “shock” (choc) in her psychiatric clinic in France, around 1926, she was referring to physical treatments meant to surprise or disturb the body and brain, restoring the mind to equilibrium. Pascal considered mental illnesses to be a kind of allergic reaction, and she argued that resetting the brain and nervous system would be necessary for recovery. Pascal’s first shock therapies, in the late 1920s, used various means to prompt a reset, including induced fevers. Around the same time, in 1927 Berlin, the neurophysiologist Manfred Sakel was developing a new form of shock treatment in which a patient received potent injections of insulin. At proper doses, Sakel found he could induce a grand mal seizure as the patient came out of a hypoglycemic coma.

Insulin coma therapy produced astonishingly positive results for mood disorders—sometimes. (One famous recipient was the storied mathematician John Nash, who received insulin shocks for schizophrenia in 1961.) But results were markedly uneven. Ladislas Meduna, the Hungarian neuropathologist, shared Pascal’s interest in the possibilities of shock therapies and came to wonder whether seizures might offer a better route to healing.

In autopsies of epileptics, Meduna observed that the brains showed “tremendous changes” that were “just the opposite of those found in schizophrenia” (then a broad term that could encompass depression, schizoaffective disorders, and much more). In particular, Meduna noted that the glia—supportive cells sometimes called the janitors of the brain, whose duties include tending to our neurons—were overproduced in the brain tissue of epileptics, and vastly underrepresented in the cranial autopsies of patients with advanced mood disorders.

Pursuing this apparent antagonism between seizures and psychosis, Meduna introduced a new shock therapy in late 1934, injecting a catatonic patient with camphor, the compound then used in mothballs. Three weeks and five camphor-assisted seizures later, the patient was moving and talking—two things he hadn’t done in four years. Meduna repeated this apparent success with four subsequent patients, sometimes replacing camphor with the synthetic compound pentylenetetrazol, branded as Metrazol or Cardiazol, which achieved a similar convulsion but faster. It was also safer than insulin comas, which carried a real risk of death.

Still, Metrazol had severe drawbacks. One of the biggest was that the chemical seemed to induce feelings of absolute terror in patients before they finally fell into the coma. Not ideal for subjects already managing profound mental disturbances.

Meanwhile, in Italy, they were beginning to practice a new form of shocking the system, one that worked faster and more effectively than insulin or Metrazol—even if no one quite knew how it worked in the first place.

At the Clinic for Nervous and Mental Disorders in Rome, a little after 11 a.m. on April 11, 1938, the neurologist Ugo Cerletti placed a pair of electrodes on the temples of a man identified as “Enrico X.” Cerletti and his colleague, Lucio Bini, used a rubber band to hold the electrodes in place and electrocuted Enrico over a period of months, around 11 times. By the end of his electro-treatments, Signore X, the first ECT patient in history, was apparently recovered. The next month, in a presentation at the Royal Academy of Medicine in Rome, Cerletti and Bini administered an electro-treatment to a patient before an audience. In 1939, Cerletti and Bini coined a verb for what they’d invented: zapare, literally meaning “to zap.”

Even from the start, Cerletti was uneasy about the awesome power he commanded with those two electrodes. “When subjecting unconscious patients to such an extremely violent reaction as these convulsions, I had a sense of illicitness and felt as though I had somehow betrayed these patients,” he recalled later. Though Cerletti didn’t dwell on it, there was also the small matter of Enrico X’s relapse. He ended up readmitted to a psychiatric institution two years later.

Prehuman trials of electroconvulsions give a taste of the freewheeling sensibility that led to the development of electroshock. For example, in order to determine whether epilepsy affected the hippocampus, Cerletti as early as 1933 had sought to induce epilepsy in dogs. To do so he placed two electrodes on each dog, one inside the mouth, the other inside the rectum.

In Shock Therapy, their comprehensive 2007 history of ECT, Edward Shorter and David Healy report with dry understatement: “There was nothing innovative about this as such, and the mortality of the procedure was very high because the current passed through the dog’s heart, disrupting its electrical (pacemaker) activity.” It was an ignoble end for the dogs. For future experiments, Cerletti and Bini began placing the electrodes on a dog’s head, adequately producing a grand mal seizure.

Later, members of the Italian team visited a slaughterhouse in Rome, where dual-electrode stunning was used to render pigs unconscious before slaughter. Bini reported that after the shock, the stunned pigs reliably experienced grand mal seizures. This was precisely the seizure response that the clinicians sought in humans, and, barring extraordinarily high voltages, it carried no threat of death. (That came later, when the meat packers slit the pigs’ throats.)

ECT COMES OF AGE

Simply reading about dog rectums and pigs for slaughter, I quickly grasped the enduring stigma associated with ECT from these two details alone.

In a few of the books my doctors had recommended, I read how insulin, Metrazol, and ECT jockeyed for primacy during a period of remarkable change in neuropsychiatry, when physical treatments challenged and then overtook the formerly dominant psychoanalytic model. Then the beginning of World War II cut supply lines, making insulin prohibitively expensive or unavailable in some areas of the globe; ECT was cheaper, safer, and clinically far more attractive than either insulin or Metrazol. The war also emptied parts of Europe, notably Vienna, of many of its psychiatric eminences: Freud moved to England, Meduna went to Chicago, and the Berlin-born psychiatrist Lothar Kalinowsky, who worked on ECT in Rome under Cerletti, decamped to spread the electric gospel in London and then New York.

ECT, then, much like gin or the novel, was born in mainland Europe, refined in England, and naturalized in the States. Many American psychiatrists welcomed the practice. “Coming at a time when war is subjecting the population of the whole world to those intolerable mental strains that precipitate mental disease,” the Science News Letter wrote with excitement in 1940, “this new use of electricity for mental health instead of for death is being enthusiastically welcomed by the medical profession.”

That was the year ECT took America by storm. “In the summer of 1940, the East Coast must have literally buzzed with electricity,” Shorter and Healy write in Shock Therapy. And those who recovered often evangelized. The procedure involved some discombobulation, many patients said, perhaps some memory loss. But it had been undeniably worth it.

At least that was the case for the success stories. There were many others. The asylum era was still chugging along in the 1940s, and asylums were justly feared and hated. Among the inmates, ECT became one more blunt tool for asylum keepers to overuse, at a time when patients had precious few rights. As with fever therapy and hydrotherapy (repeated, prolonged, and often nonconsensual dousing), some asylum doctors in the 1940s and ’50s seem to have taken on ECT as a challenge: How much of this can my patients stand? Bini himself advanced an argument for the metodo dell’annichilimento, or “annihilation therapy,” which required zapping a patient multiple times a day. Other doctors picked up this method with the express purpose of rebuilding the patient’s personality from scratch.

These were tantalizing and terrifying things to read when I was deeply depressed and considering submitting to ECT myself.

THE MODERN, SOFTER SIDE OF ECT

My doctors told me that today’s ECT was quite different. No one was doing annihilation therapy in the U.S. anymore, one said. Plus, modern or “modified” ECT was much more civilized because the electrical pulses were briefer and more targeted, and the doctors gave you anesthesia and muscle relaxers and even an oxygen tube to prevent asphyxia. All the horror stories, my psychiatrist said, dated from the era of “unmodified” ECT. “We’re past the grisly stuff now,” he told me.

But that didn’t make the history any easier to read.

Yes, anesthesia and tranquilizers entered the picture—particularly in 1951, with the introduction of the muscle relaxant succinylcholine, still used in ECT today. But many asylums and hospitals were slow to adopt these new, gentler practices. According to the science journalist and ECT skeptic Robert Whitaker, more than 1 million patients got forced ECT in the 1940s and ’50s. Even kids didn’t escape the electrodes. From 1942 to 1947, Bellevue Hospital in New York treated 98 children, some as young as 4, with a long sequence of ECT treatments.

My heart rate slowed a bit when I read about the softer side of ECT. It’s considered sufficiently gentle to be offered as a first resort for elderly patients who can’t tolerate SSRIs (selective serotonin reuptake inhibitors) or other pharmacological interventions. I was even more surprised to learn that ECT didn’t attain its stigma—its association with persecution and alleged brain damage—until the mid-1960s, thanks in part to emerging psychopharmaceuticals but mainly amid a frantic new campaign to abolish ECT.

THE CAMPAIGN AGAINST ECT

The antipsychiatry movement was a coalition between heterodox psychiatrists and members of the New Left. It emerged at a time when the antiauthoritarian ethos of young baby boomers came into conflict with a procedure that recalled the tyranny of asylum medicine. “Shock resonated because, like nuclear energy, [ECT] had the aura of an out-of-control technology imposed by an authoritarian society,” observed the journalist Larry Tye, who, with Kitty Dukakis, authored an illuminating 2006 book about electroshock. By the 1970s, Tye reported, detractors had begun “picket[ing] outside electroshock clinics with signs linking the practitioners to Spanish inquisitor Tomás de Torquemada and Nazi executioner Adolf Eichmann.” The anti-ECT campaign has been vigorous and well funded, thanks not least to munificent contributions from the Church of Scientology, which has also played a central role in organizing.

Accordingly, ECT therapy was severely curtailed in most parts of the country. Numbers are tricky, as there’s no central trove of data, but the most plausible estimate is that the yearly number of ECT patients in the U.S. fell from around 300,000 in the 1950s and ’60s to around 59,000 by 1975; by 1980, that number had dropped further, to around 32,000. “It is as though penicillin had suddenly gone into a decline in the absence of any objective circumstance that might discourage physicians from prescribing it or patients from accepting it,” Shorter and Healy wrote.

One can understand the skepticism. I know I did. But as I read about the campaign to expunge ECT, I wondered whether the problem was a failure to distinguish the baby from the bathwater. It seems statistically undeniable that, during ECT’s period in the wilderness, millions of depressed or bipolar or schizophrenic Americans were functionally denied recourse to ECT. One doctor told me we’d never know how many died because ECT was drummed out of so many American hospitals. ECT’s frequently lurid portrayals in popular culture only bolstered the eye-popping claims of the antipsychiatry camp: ECT turns you into a zombie; it erases your entire memory; it will literally break your back.

I found a few books that mounted a defense of ECT. Sometimes half-apologetic, other times closer to triumphant. I read the Shorter and Healy history, Carrie Fisher’s memoir about electroshock, and Tye’s book with Dukakis, entitled Shock: The Healing Power of Electroconvulsive Therapy. (“In some ways ECT is less traumatic for me than going to the dentist, and certainly less frightening than the root canal,” Dukakis wrote.)

But when I tried to find some authoritative consensus about memory loss, I kept meeting a chorus of contradiction from the medical literature. Some doctors said memory loss would be mild; others said it could be “more pronounced” (than what?); but the third and loudest group said that memory loss was a canard, a talking point for antipsychiatrists, adding that most patients reported few to no cognitive problems. When I went online to eavesdrop on patient forums, the consensus was clearer: Memory loss is very, very real.

I’m only speculating here, but I do wonder whether, in their zeal to defend a lifesaving treatment from demagogic skeptics, some ECT advocates may have pushed back with their own immoderate claims. Whatever the cause, it’s a strange experience to have different doctors tell you different things about what’s about to happen to your memory. It’s even stranger when, after a few treatments, you can’t quite remember which doctor told you what.

GENERATING NEW NEURONS (AND OFTEN HOPE)

The most baffling, endlessly intriguing questions about ECT are how and why it works. Physicians since ancient times have recorded the sometimes salutary effects of seizures on certain forms of mental illness, yet even today we don’t know precisely why a given brain is happier, or clearer, or calmer, following a few seizures. There’s an old saw comparing ECT to aspirin: We don’t know how it works, but it works magnificently. Still, a few general ideas about the mechanism of ECT are widely accepted.

During the seizure, affected neurons all fire at the same time, several times per second. Some doctors refer to this effect as the brain “resetting” itself. As a result of this firing, the brain releases a welcome surplus of neurohormones such as norepinephrine and neurotransmitters such as dopamine. The most intriguing recent account for ECT’s high rate of response—often reported at around 75% to 80% for major depressives who’ve failed to find relief with pharmacology—is that the release of these brain chemicals can stimulate neural growth across various regions of the brain. Per brain imaging, just a few ECT treatments are often enough to spur visible volume increases in multiple cortical regions, not least in the hippocampus, the area of the brain that is prey to atrophy during major depression and other maladies, including addiction, certain schizoaffective disorders, dementia, and Alzheimer’s.

ECT’s demonstrated neurogenetic properties offer a compelling explanation for why the treatment remains a most effective antidepressant. That the brain grows and shows markedly more, and more regular, activity after ECT treatment is potentially confusing to lay­ people, many of whom take the treatment to be inherently damaging—­“scarring” is the preferred participle among the skeptics, as though doctors have applied a knife or a white-hot poker to your brain. Of course, surprisingly little electricity goes through you; not enough to singe or scar. Comparisons are tricky, because the human machine and, say, the refrigerator require different currents, but the amount of energy that zaps you during a single ECT session is mercifully modest—perhaps the equivalent of charging your phone by 1% or powering a lightbulb for a few seconds.

Discoveries about hippocampal growth after ECT are straightforwardly exciting. For those of us who grew up in the 1980s and ’90s, the essential unrecuperability of brain cells and memories was treated as a basic fact in “just say no”–style antidrug campaigns and even biology class. That was all wrong, even then. Neurogenesis had been established in small mammals and birds by the 1980s, while in the 1990s, we discovered that the human brain has its own stem cells—pluripotent little guys that can become new neurons, sometimes creating new connections that support neuroplasticity, as well as fresh glial cells to clean and maintain the brain’s circuits.

Still, certain disagreements persist. Some of the keenest students and proponents of ECT believe the seizure is the crucial mechanism, while other experts argue it’s the body’s postseizure response that hits the reset button, rather than the seizure itself. If the latter group is right, that opens the intriguing possibility that future advances in brain imaging and brain stimulation might reveal a gentler shortcut toward triggering an anticonvulsive response, hypergenerating these chemicals, and reinvigorating the hippocampus and its neighbors, all without the dreaded seizure.

Regardless, ECT remains the nuclear power of psychiatric treatment: much maligned, often misused, yet sturdy and powerful, the gold standard that we’ve somehow disowned. Among latter-day electrical treatments, which include transcranial magnetic stimulation and vagus nerve stimulation, none thus far has shown the efficacy of modern ECT. Particularly for acute cases—people who might not have the time to wait for another drug to work or not to work—ECT is unmatched.

It could also be on the verge of becoming much more effective. Cutting-edge research into ultrabrief pulses and hypertargeted forms of the treatment offer the tantalizing possibility that future patients might receive greater benefits with fewer side effects. But we’re not there yet.

There are, of course, legitimate lines of skepticism. Some skeptics raise reasonable methodological points. For example, while we can draw on various controlled trials into ECT’s efficacy, placebo-controlled trials—in which a control group undergoes a sham version of the treatment—are rare, largely because they raise major ethical concerns. (Fooling a depressive into thinking he’s had a seizure is both tricky and mean.)

The other major concerns involve consent and side effects. In the U.K., for example, there are complaints that some older patients, especially women, have received ECT against their will, and that doctors have minimized side effects, leaving patients shaken when they discover unexpected memory deficits.

Today, the anti-ECT movement is gaining ground again. In the U.S., opponents have once again taken to the courts, where litigation over undisclosed risks helped push one manufacturer of ECT machines into bankruptcy in 2021. A separate lawsuit produced a 2026 settlement from two psychiatrists accused of understating potential harms.

As a journalist rather than a researcher, all I can say with certainty is what I’ve seen empirically in myself and other patients: ECT can have miraculous-seeming positive effects for the deeply depressed. It also verifiably inhibits the consolidation of more recent memories. My treatment was voluntary, and I lost access to more of the previous months than I’d been prepared for. What I got in exchange, though, was real healing; slow at first, then undeniable. Without ECT, I could’ve harmed myself and continued courting death. Instead, I returned to life in all its mess and splendor. Historically, ECT has indeed been much abused. But among the many ways to abuse ECT is withholding it from a patient whose life it could save.

Excerpted from Jolt: My Electric Journey Out of Darkness by Ted Scheinman. Published by Simon & Schuster. Copyright © 2026 by Ted Scheinman. All rights reserved.

This Googlebook is driving me crazy

1 week ago

Right now, Googlebooks make more sense for Google than they do for users.

After trying for years to bring Android features to Chromebooks, Google decided to start over with Android as the foundation for a new laptop operating system. The first wave of Googlebooks are high-end laptops (priced $900 and up), with Google pitching them as the ultimate computers for Android enthusiasts.

But in their current form, Googlebooks are a disjointed mess, with websites and apps that don’t work quite right and AI features that aren’t fully fleshed out. While having a single operating system and app platform across laptops and mobile devices might make life easier for Google, it doesn’t yet translate to a better experience than the Chromebooks that Google aims to replace.

Android’s not ready for laptops

For the past month I’ve been using Acer’s Googlebook 14, which is the cheapest Googlebook at $900, and the only one whose touchscreen flips around into tablet mode. I have no major complaints about the hardware, which has a featherweight 2.51-pound chassis, a rich 120 Hz 3K OLED screen, and a smooth trackpad (albeit a “diving board” type whose click mechanism stiffens at the top edge).

My frustrations are more on the software side, where Android isn’t ready to carry a real laptop workload.

While Google is pushing developers to optimize their Android apps for large screens, keyboards, and trackpads, Google itself hasn’t always bothered to do so. The preloaded version of Gmail, for instance, is not an optimized Android app, but rather a web-based desktop app powered by the Chrome browser. Same goes for Google Docs, Drive, Meet, and YouTube. Tapping the menu button in these apps reveals a “Running in Chrome” disclaimer, and any Chrome extensions you install will apply to these apps as well.

This wouldn’t be a major concern if the web apps worked the way they’re supposed to. In Gmail, though, cursor hover controls don’t work, so you can’t quickly trash or archive messages line by line in the inbox. In Google Drive, documents open in a Chrome browser tab, while the same documents open in their own windows via Google Docs or Sheets. And if you try to select text in Google Docs using the touchscreen, the comment button pops up directly over the right text selector, making it unreachable.

Third-party Android apps can also have issues. Signing into Slack bounces me out to a Chrome tab, which then fails to authenticate back inside the Android app, so the only way to gain access is by manually entering a URL and password for each workplace. Even then, Slack’s Android app is poorly optimized, without the desktop version’s persistent sidebar.

Video apps still feel unfinished on Googlebooks as well. Those that Google deems “optimized for desktop,” such as Netflix and Disney+, will automatically open in full screen, which is jarring but at least fits the use case of watching video. Others, such as Fox One and NFL, open in a floating desktop window, and clicking their full screen icons accomplishes nothing. You’re actually supposed to ignore that button, highlight the maximize button in the window controls, then click the little “Full Screen” icon. Optimized or not, none of these apps support always-visible picture-in-picture mode like their mobile counterparts.

One last nitpick, admittedly a niche one: Bookmarklets don’t work on Googlebooks. I keep more than dozen of these in my browser’s address bar, which allows me to launch site-specific searches, create page excerpts, find archived versions of web pages, and remove “sticky” page elements. Not being able to use them on the Googlebook is a reminder that you’re not quite getting a full Chrome desktop experience.

The app situation will likely improve over time. When I started using the Googlebook, it wouldn’t let me install the Bluesky app and directed me to the web version. Now the app is available. Google Docs has also gained the ability to mark individual documents for offline use, which wasn’t possible at the outset. The Google Play Store has a tiered labeling system—“Optimized for Desktop,” “Compatible with Desktop,” “Better on the Website”—so you know what to expect, but for now the ecosystem is a work in progress.

Android synergies in progress

Googlebooks do have some neat features for Android users, the best of which is being able to access your phone’s files directly through the laptop’s Files app. There’s also “cross-device continuity,” which doesn’t work with many apps, but lets you quickly continue editing a message in Gmail or a document in Google Docs after opening it up on your phone.

App casting is rougher. If you have a phone running Android 17, this lets you view your phone’s apps through a Googlebook without installing them, and while this might be useful in some scenarios, it’s a subpar experience with jerky scrolling and occasional formatting issues. It can also lead to annoyances, like when a phone notification pops up on the Googlebook, and tapping it takes you into the janky app-casting version instead of the one that’s installed on your computer.

Android as a desktop operating system will also inherently pose some limitations. AI apps like ChatGPT and Claude lack the device control and folder-based file permissions that their desktop apps offer, and there are no command bar apps akin to Raycast for Windows and MacOS. The relative simplicity of an Android-based laptop may be a positive for some, but for me it’s a hindrance.

Too many Geminis

Beyond just catering to Android enthusiasts, Google is pitching Googlebooks as a premier surface for its Gemini AI assistant. In practice this means there are four main ways to invoke Gemini from the computer:

  • Wiggling your mouse cursor
  • Saying “Hey Google”
  • Clicking the Gemini icon in the dock
  • Clicking the Gemini icon in the Chrome taskbar

While having many entry points to Gemini makes sense, the execution feels disjointed. If you open the full Gemini app while a pop-up chat window is already open, for instance, the chat window closes and the full app cuts off your conversation, switching to a new one. The Chrome sidebar and pop-up window are also separate, so conversations in one don’t carry over to the other.

Those little inconsistencies tripped me up constantly, as I’d have to keep careful track of which conversation was happening where based on how I’d originally invoked it. Instead of just cramming AI touch points into different parts of the operating system, Google needs to do a better job of tying them all together.

There’s also a missed opportunity for Gemini to act more as an AI agent. While desktop AI apps on other platforms can work directly with files and folders on your computer, that’s currently not possible with Gemini on Googlebooks.

What’s next for Googlebooks?

Even if Googlebooks don’t feel ready now, they carry an air of inevitability. Google has already indicated that it will phase out Chrome OS by 2034, and it has outlined some vague plans for existing Chromebooks to turn into Googlebooks. Presumably that means Googlebooks will move down-market over time to include the kind of lower-end machines that defined the Chromebook business.

In the meantime, the question is whether Google will sufficiently invest in Googlebooks to address their weaknesses, instead of letting the project languish as it so often does (see: Stadia, Reader, its many attempts at social networking). One reason GoogleBooks feel so unpolished right now is because of Google’s hot-and-cold relationship with Android tablets, leading to a lack of app developer buy-in on devices with larger screens. With Google, there’s always a risk that it gets distracted by the next shiny object and leaves important work undone.

You can see how Googlebooks might eventually make sense: Combine a desktop-class Chrome experience with true Android app support, with easily available AI to help with research and automation. Maybe Google will get there someday, but first it needs to get more of the basics right.

Rian Johnson’s new camera app shows what’s missing from iPhone photos

1 week ago

Last week saw an unexpected hit on the iPhone App Store: a $12.99 camera app without a free trial or a subscription. Chromagraph Camera shot to the top of the paid Photo & Video charts after Knives Out and Star Wars: The Last Jedi director Rian Johnson revealed that he had assigned his longtime cinematographer, Steve Yedlin, to work on his dream photo app for the iPhone.

Johnson, who partnered with Yedlin on every feature film throughout his career—from 2005’s Brick to last year’s latest Knives Out mystery, Wake Up Dead Man—suggested that this was a project he’d wanted to exist for a while. 

“So I recently badgered my cinematographer @steveyedlin into making my ideal iPhone still camera app,” Johnson wrote on X. “It shoots RAW, bypasses the iPhone’s look processing and instead uses the same image processing we use on our movies.”

{"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 \u003Cem\u003EMulticore\u003C\/em\u003E\u003C\/strong\u003E\u003C\/strong\u003E","dek":"\u003Cem\u003EMulticore\u003C\/em\u003E 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":""}}

Growing discontent

The pitch speaks to a growing discontent with the way the default camera app on the iPhone handles image processing. Apple’s computational photography pipeline is incredibly good at making sure you don’t miss the moment, and at rescuing challenging exposures, blending several frames together at once so you don’t blow out highlights or lose details in the shadows. But flattening the actual world like that comes at the expense of traditional photographic character, leaving images feeling artificially sharpened and reliant on HDR.

Yedlin is uniquely qualified to rethink how a smartphone camera might turn light into a final photo. In filmmaking circles, he is well-known for his research into color science. His Display Prep Demo in 2015 was influential in showing how a particular filmic look isn’t unique to actual celluloid or a certain camera brand, but something that can be replicated by transforming raw digital sensor data. Yedlin argued that creators should be the authors of their own looks, sculpting elements of the image like grain, tonal rolloff and halation, or the glow that surrounds bright light sources on film.

Chromagraph showcases that philosophy on a smaller scale. Combined with work from color science programmer Eric Cameron, it takes the iPhone’s raw sensor data and tweaks it to Yedlin’s specific taste. You don’t see the effect while framing your shot in the viewfinder; the effect is applied after you capture the image. 

But because Chromagraph stores the original raw files, you can later tweak parameters like tone, grain, and halation after the fact without losing image quality, then export the JPEG to your camera roll. That’s a world away from traditional photo filters. Even Apple’s ProRAW format still retains elements of the iPhone’s multiframe merging.

The look won’t be for everyone. Next to a full-resolution ProRAW shot, you’re inevitably going to get a softer, grainier image that will often blow out the highlights. The app’s user interface could also use some work, as it doesn’t feel fully native to iOS. 

But in my testing, I’ve been struck by how cohesive Chromagraph feels across a variety of situations. It’s like the app turns the iPhone into a singular 35mm point-and-shoot camera with a well-chosen film stock. I’ve rarely been disappointed with the results.

“The custom look is opinionated,” Yedlin told Newsshooter. “It’s just the nuanced and idiosyncratic rendering of tonality and chromaticity (as well as grain and halation) that Rian, Eric, and I personally like. And the grading knobs allow you to make pretty dramatic adjustments while keeping within that overall core look.”

Not the first try

Chromagraph isn’t the first attempt to bring a more traditionally photographic style to smartphone imagery. For years now, Chinese phone makers Oppo and Xiaomi—in partnership with pro camera titans Hasselblad and Leica, respectively—have deployed tasteful image processing pipelines right in their default camera apps. The standard Leica Authentic mode on Xiaomi phones is particularly good; this year’s Xiaomi 17 Ultra even included an option to directly emulate the CCD sensor of the classic Leica M9.

On the iPhone side, other apps have offered their own take on stripped-down raw image capture, like the Process Zero mode in Halide. It’s also fair to say that Apple itself is starting to notice the demand; the iPhone 18 Pro comes with an updated take on the Photographic Styles feature that includes a customizable film setting that adds halation and grain. It takes a little tweaking to get right, but it is now possible to set up an iPhone to take good-looking filmic photos right in the regular camera app.

What I think speaks to people about Chromagraph, though, is that it looks even better without requiring any setup at all. It’s just one creative vision that you can take or leave. Judging by the reactions I’ve seen from countless people with whom it’s clearly resonated, you may well want to take that vision for yourself—even if it does come in the rare form of an app you have to pay for outright in 2026.

“I’ve messed with plenty of good 3rd party tools but this is an elegant one-stop-shop,” Johnson said on X. “It just takes beautiful shots with Steve’s special sauce processing.”

{"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 \u003Cem\u003EMulticore\u003C\/em\u003E\u003C\/strong\u003E\u003C\/strong\u003E","dek":"\u003Cem\u003E\u003Cem\u003EMulticore\u003C\/em\u003E\u003C\/em\u003E 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":""}}

A week after the Democrats’ blue wave, Trump’s biggest tech allies are still silent

1 week ago

“Guess my election-interfering space lasers didn’t work this time.”

That was Elon Musk’s sarcastic response on X to the blue wave in the 2026 midterm elections.

Guess my election-interfering space lasers didn’t work this time 🤷‍♂️

— Elon Musk (@elonmusk) November 4, 2026

But almost a week later, Musk remains one of the few prominent Silicon Valley figures who backed Trump and his Republican allies to have publicly acknowledged the Democrats’ sweeping victories.

It’s no surprise that Musk spoke out, given his outspoken nature in general and his massive spending on Republican candidates. The sometimes trillionaire CEO of Tesla and SpaceX spent about $474 million on the midterms, a record amount, and roughly $300 million more than the next-biggest donors, Richard and Elizabeth Uihlein, according to The New York Times.

Yet Musk walked away from the November 3 election with little to show for his money, especially in his adopted home state of Texas, where nearly half of his nationwide donations had gone toward losing Republican candidates.

As results poured in, Musk deleted his “Vote red today!” post on X, replacing it with “Data centers in space looking pretty appealing right now.”

A generally muted response

While many of Trump’s Silicon Valley allies have remained quiet about the Democrats’ victories, some tech executives have addressed what the results could mean for their businesses—particularly the AI infrastructure boom.

OpenAI CFO Sarah Friar said the company would continue investing in AI data centers, even after voters rejected local measures supporting their construction in 40 out of 41 instances, CNBC reports. “When I think about the midterms, I need to keep thinking about how do we keep investing to meet that customer demand,” Friar said on November 6. “It starts early with having built relationships with all of the hyperscalers, and it will continue to mean that we raise capital.”

The day before, OpenAI announced a new commitment to pay for its own data centers, protect local resources, issue impact reports, and create local economic opportunities. The company billed the initiative, seemingly designed to ease concerns among communities and policymakers wary of AI infrastructure, as “Setting a new standard for responsible AI infrastructure.”

Meanwhile, Box CEO Aaron Levie told CNBC the results could raise questions about where data centers get built and how AI supply chains operate.

But on the broader political implications of the election, many of Trump’s Big Tech allies have remained mum. Meta CEO Mark Zuckerberg, Alphabet CEO Sundar Pichai, and Nvidia CEO Jensen Huang—all past Trump donors—have yet to comment publicly on the results.

The silence marks a change from 2024

When Trump won the 2024 presidential election, many of those same tech leaders were quick to offer congratulations, often on Musk’s X.

“Congrats to President Trump,” OpenAI CEO Sam Altman posted at the time. “I wish for his huge success in the job. It is critically important that the US maintains its lead in developing AI with democratic values.”

Pichai wrote, “Congratulations to President @realDonaldTrump on his decisive victory. We are in a golden age of American innovation and are committed to working with his administration to help bring the benefits to everyone.”

Apple’s then-CEO Tim Cook echoed that sentiment: “Congratulations President Trump on your victory! We look forward to engaging with you and your administration to help make sure the United States continues to lead with and be fueled by ingenuity, innovation, and creativity.”

“Congratulations to President Trump on a decisive victory,” Zuckerberg wrote on Meta’s Threads. “We have great opportunities ahead of us as a country. Looking forward to working with you and your administration.”

Some Silicon Valley leaders welcomed the results

Not everyone in Silicon Valley has been so quiet. While the tech industry has shifted sharply to the right in recent years, several prominent figures, particularly those who backed Democrats, welcomed the results.

Take Democratic donor Reid Hoffman, cofounder of LinkedIn. “Voters made clear they demand a different kind of country,” Hoffman wrote on X. “Americans want an economy that can compete globally and deliver for them at home. And they made clear that they want leaders they can trust to fight for their constituents.”

1/ A few thoughts on tonight’s election:
Voters made clear they demand a different kind of country. Americans want an economy that can compete globally and deliver for them at home.
And they made clear that they want leaders they can trust to fight for their constituents.

— Reid Hoffman (@reidhoffman) November 4, 2026

Hoffman donated about $16.5 million to James Talarico’s successful Senate campaign in Texas, according to The New York Times. He also gave nearly $9 million to the Democratic Party of Wisconsin, where Democrat David Crowley won the gubernatorial election.

Then there’s Lane Kasselman, the co-CEO of Blockchain who previously worked on Hillary Clinton’s campaign. “Contrary to how it’s portrayed in the media, people across the political spectrum have a lot more in common than they might think,” Kasselman wrote on X. “Whether you’re from my home state or my adopted state, most Californians and Texans aren’t that different: we love our kids, like clean air, want prices to come down and think Mexican food is awesome.”

I spent my early career working in Democratic politics in California, and today I co-lead a company based in Texas. Contrary to how it's portrayed in the media, people across the political spectrum have a lot more in common than they might think. Whether you’re from my home state…

— Lane Kasselman (@kasselman) November 4, 2026

Kasselman continued: “I’m proud to see pragmatism win in Texas. And with the most pro-crypto Congress in history, there’s a real opportunity to build on that pragmatism in Washington.”

Virgin Group founder Richard Branson, meanwhile, criticized the election spending of his fellow billionaires in a LinkedIn post: “As a successful entrepreneur, I have felt embarrassed by the billions that tech entrepreneurs and hedge fund managers have spent on—mostly Republican—campaigns this election cycle, in an attempt to buy influence and control over public policy.”

The election could have significant consequences for Silicon Valley, particularly as Democrats take control of Congress and questions mount over the AI industry’s enormous infrastructure ambitions. But while the industry’s Democratic supporters have been quick to celebrate, many of the tech leaders who enthusiastically welcomed Trump’s return to power have had little to say about his party’s sweeping defeat.

The Jensen Huang spy scandal is becoming a White House problem

1 week 1 day ago

The scandal engulfing Nvidia CEO Jensen Huang is no longer simply a question of whether one of Silicon Valley’s most powerful men had an affair with a suspected Chinese spy. It’s also a question of what the White House knew—and what it might have chosen to overlook.

Huang and Nvidia confirmed the relationship Sunday, following a bombshell New York Times report. Alice Huang (no relation to Jensen)—the woman at the center of the revelations—is an executive at Singapore-based data center company Megaspeed, a major customer of Nvidia chips. U.S. intelligence surveillance reportedly showed that she had been funneling information directly to the Chinese government.

What remains unclear is how much sensitive information may have been compromised, what Huang knew about the woman he was seeing, and why U.S. intelligence officials apparently failed to warn him about her suspected ties to Beijing. The answers could prove deeply damaging to both Huang and the White House.

“It was pretty clear to a lot of people that [Jensen Huang] would be a target of Chinese intelligence collection,” given his proximity to President Donald Trump and his leadership of a company that is “central to the U.S.-China AI competition,” says Sam Bresnick, a research fellow at Georgetown University’s Center for Security and Emerging Technology.

Nvidia’s changing story

The admission of the affair also contradicts how Nvidia and Huang himself had previously characterized his ties to Alice Huang. As recently as October 29, company spokesperson John Rizzo described their interactions as routine business dealings, while Huang insisted they were “completely normal and not unusual.”

Huang now faces an investigation by Nvidia’s board, although directors said in a statement Sunday that they had “no reason to believe that either national security or Nvidia secrets were compromised” and expressed “confidence in Jensen and his leadership team.”

The board’s investigation may be the least of Huang’s worries. The powerful CEO faces the prospect of bipartisan congressional scrutiny, along with mounting questions about his influence in Washington and his future at Nvidia.

The questions Congress wants answered

Late last week, three House members—Republicans Brian Mast, John Moolenaar, and Eric Crawford—sent Huang a letter requesting that he attend a closed-door hearing on the matter in Washington, D.C. Asked by Fast Company on Monday whether Huang intends to attend, Nvidia declined to answer on the record. Next year, the new Democrat-controlled Congress will likely call a number of tech leaders, including Huang, to testify about their dealings with Trump.

On Monday, Senator Chris Coons of Delaware called for an investigation into Huang and Nvidia to determine whether national security had been compromised. “We need answers about the relationship between AI billionaires and the Trump administration,” he wrote on X. Coons is poised to chair the Senate Foreign Relations Committee when Democrats take control of Congress in January, giving him considerable power to pursue those answers.

There is certainly plenty to investigate. For one, we don’t know whether Alice Huang obtained sensitive information from Jensen Huang and transmitted it to the Chinese government. Bresnick says Huang and his devices may have been in possession of several types of valuable information, including trade secrets about upcoming Nvidia chips, the company’s research strategies, or U.S. government plans to restrict China’s access to advanced AI technology.

“Did he provide information to her about the Trump administration’s thinking on AI—U.S.-China competition, export controls, and trade policy more broadly?” Bresnick asks. “Because if [the Chinese] have greater texture on what the administration’s thinking, it’s easier for them to position themselves.” (The White House did not respond to Fast Company’s request for comment.)

It’s also unclear what Huang knew about the woman, and when. He certainly knew he was having an extramarital affair with an executive at a customer company, which violates Nvidia’s code of conduct. But we don’t know when he learned that she was a suspected spy.

What did the White House know?

Then there’s the question of when U.S. intelligence learned about Alice Huang’s activities.

According to the reporting, the U.S. government had known about her ties to Beijing for months. Bresnick finds it strange that intelligence officials apparently failed to warn the Nvidia CEO, given his close relationship with the president and his position atop a company developing some of the world’s most advanced AI technology. “The fact that this potentially could have been at risk” makes the apparent lack of intervention all the more puzzling, he says.

One possibility is that the White House prevented intelligence agencies from warning Huang about the suspected spy. After all, the Trump administration has compelling reasons to keep Huang and Nvidia out of the headlines.

The stock market’s strong performance has been a central selling point for the administration, and Nvidia is the crown jewel of the AI-related stocks that have driven much of its growth. AI and AI-related companies accounted for as much as half of the market’s gains this year, according to a new J.P. Morgan report.

The White House may have had an even more direct financial incentive. The federal government stands to collect revenue from Nvidia’s AI chip sales to China. Last year, Huang persuaded Trump to allow Nvidia to sell its H200 chips in China, provided the U.S. government received a 25% cut of the sales.

Huang also reportedly discussed selling Nvidia’s top-of-the-line Blackwell graphics processing units in China, but that proposal was ultimately shut down over national security concerns. In August 2025, Huang agreed to give the U.S. government a 15% cut in exchange for permission to sell the lower-performing H20 chip to China.

“Was the government hoping to get a bunch of money from these deals, from Chinese buyers getting access to these things?” Bresnick asks.

Nvidia’s influence comes under scrutiny

Alice Huang’s employer, Megaspeed, became an official Nvidia Cloud Partner in November 2023 and purchased more than $4.8 billion worth of Nvidia chips over roughly two years, according to the Associated Press. But the company was also under investigation by the U.S. Commerce Department over suspicions that it was funneling restricted AI chips to China.

A New York Times investigation last year found that Megaspeed’s offices in Singapore and Malaysia were largely empty, while many of its executives are based in China. Commerce Department officials who spoke to The Times even described finding boxes of restricted Nvidia chips in a Megaspeed facility in Malaysia.

On Monday, The Times reported that the Commerce Department’s investigation came to an abrupt halt last year after Nvidia’s general counsel, Tim Teter, contacted investigators seeking information about the case. Teter reportedly told them that Jeff Kessler, a senior Commerce official appointed by Trump, had given Nvidia permission to reach out. The communication alarmed investigators, some of whom worried that Kessler’s involvement had potentially compromised their work.

The case was subsequently transferred to the Justice Department, which is now conducting a criminal investigation into Megaspeed. Officials familiar with the matter told The Times the transfer was intended to prevent the investigation from being compromised.

There is no evidence that Kessler instructed investigators to share information with Nvidia, but the unusual circumstances of the transfer have raised questions about the company’s influence over the very officials responsible for policing its chip sales.

For Bresnick, the investigation also raises questions about how Megaspeed was able to acquire billions of dollars’ worth of advanced Nvidia chips despite suspicions about its operations. “We had no idea where those chips went,” he says.

Together, the revelations raise questions about whether Nvidia’s political influence complicated the government’s response to potential national security threats. This all comes at a politically sensitive moment: Corruption was a major factor in voters’ decisions in the midterms, and Nvidia now risks becoming entangled in a controversy that could damage its standing in Washington.

Despite Huang’s scandal, Creative Strategies chairman and longtime analyst Tim Bajarin suggests those calling for his ouster could meet serious opposition. 

“Removal of Jensen would be a very difficult decision because he is the heart and soul of Nvidia,” Bajarin says. “I have met with Jensen many times over the years, and he was always the smartest guy in the room. More importantly, he has the global vision for Nvidia’s future and needs to be the one driving their future.”

This tiny government division charged with AI safety is now reporting to the White House. Experts are concerned

1 week 1 day ago

With public anxiety growing over mounting reports about out-of-control artificial intelligence agents, the tiny U.S. government division charged with assessing the safety of new AI offerings lacks a permanent leader and has stopped publishing its research.

Officials with the Center for Advancing Innovation and Standards for Super Intelligence, or CAISSI, an independent division of the Commerce Department, recently were asked to informally report to an office of the White House. That’s according to two former government officials with knowledge of the ongoing conversations who spoke to The Associated Press on condition of anonymity to avoid repercussions for current staffers.

The new arrangement concerns some critics, who worry that the executive branch might be politicizing decisions that could affect everything from cybersecurity to bioweapons and that the center’s findings could be altered or muted in deference to the tech industry, which has been largely entrusted to police itself.

“These are really powerful institutions that can drastically improve our lives, or the alternative. And the hope is that you can put in nonpartisan or bipartisan technocratic experts to enact policy in a way that best serves the American public,” said Caroline Siegel Singh, an AI and emerging technology program manager for the Federation of American Scientists.

Others argue that bringing AI safety officials closer to the orbit of President Donald Trump, who has taken an outwardly hands-off approach to the issue, could give the division more prominence — and perhaps more funding.

The center, which has experienced departures and until recently was called the U.S. Center for AI Standards and Innovation, is allocated $10 million per year. That’s about one-eighth the annual budget of its counterpart in the United Kingdom, which is far less celebrated for its frontier technologies.

In April, the division’s director was hired and then fired within days. His replacement lasted a few months before he resigned, and the center has been without a permanent leader ever since, although the head of its parent institute doubles as its acting director.

In August, Congress started asking why the center suddenly stopped making public its reports about how safely certain AI systems were performing. San Francisco giant OpenAI also had chimed in by then, asking policymakers to throw more money and support behind the division.

The White House didn’t respond to specific questions about CAISSI. But it said in a statement that Trump had formally established his new Super Intelligence Force — the president’s preferred term for artificial intelligence. Trump has commissioned a report on risks from his Council of Advisors on Science and Technology, and will ensure “continued U.S. leadership on SI for the benefit of the American people,” the White House said.

An industry that polices itself

The federal government often takes an active role in deciding what is safe for the public.

Take the automobile industry, for example. The National Highway Traffic Safety Administration sits crash test dummies in cars and then runs them into walls to measure how human bodies are affected by vehicle collisions. Government researchers then analyze the crash test data, share it with the public and prescribe national standards to improve vehicle safety, all with the mission of saving lives and lowering economic costs from crashes.

With AI models, safety testing looks very different.

Under then-President Joe Biden, leading AI companies voluntarily agreed to promote “safe, secure, and transparent development” of their models by sharing details with the newly created division of the Commerce Department and following the administration’s guidelines for tests and transparency reports.

In September, following alarming reports of AI systems eluding their makers’ control, Trump asked leading AI companies to undertake “tremendous self-policing” as part of a voluntary safety pact, but his administration hasn’t said how it will monitor the AI companies’ implementation.

The technology is advancing so rapidly that many models are now deployed instead as AI agents that work without human prompting and can plan out myriad steps in real-world environments, such as in a recent case in which an AI agent sent a false tip to Philadelphia police about an unsolved homicide case.

As AI capabilities have grown, the government’s capacity to independently check these systems hasn’t kept pace. There still is no standard process for testing how safely AI agents are acting in the real world, even in simple cases such as when AI agents book a flight, said Sayash Kapoor, an incoming assistant professor at the University of California, Berkeley School of Information.

“Our previous methods would just look at the final answer, the final check, whether the flight was correctly booked. And in this case, it might seem as if the task was solved correctly,” said Kapoor, the co-author of a new paper examining agent evaluations. “But unless you’re looking at the entire set of steps that an agent took, you won’t be able to tell that the agent made a huge mistake.”

Trump has prioritized dominance in AI

When Trump started his second term, he initially empowered venture capitalists David Sacks and Sriram Krishnan as his AI czar and senior AI policy advisor, respectively. Their expertise was in leading tech investments, not the technical evaluation of AI models.

Both stepped down earlier this year but Sacks now serves on Trump’s technology advisory council. And after a relatively unknown Anthropic engineer resigned, warning about the dangers of AI development, Trump made a series of policy changes.

Trump empowered National Intelligence Director Jay Clayton to lead his new Super Intelligence Force, which the president said would keep the U.S. dominant in AI development.

Clayton’s appointment to the force, along with the Pentagon’s chief technology officer, Emil Michael, shows how the administration is pushing to frame decisions around model safety as a national security matter that needs executive discretion and secrecy, overshadowing AI’s risks to consumer protection, privacy, and civil rights, said Amos Toh, a senior counsel at the Brennan Center for Justice.

Even as Silicon Valley companies have disclosed their models’ involvement in troubling cyberattacks this year, CAISSI has kept a low profile when it comes to testing for AI safety.

Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman also have been promoting the work of independent evaluation firms to help vet their companies’ safety practices, but the AI giants can limit their access and dispense with their evaluations.

“There’s this unclear mission for government evaluators if the largest AI companies are also being encouraged to set up their own internal evaluation systems and/or hire external evaluators,” said Logan Koepke, a senior project director at the civil rights and technology nonprofit Upturn. “Like, what role does government testing actually play in that scenario?”

Letting the labs police their own models and failing to clarify the role of government in AI safety testing weakens independent checks on AI systems that can cause real-world harms, Toh said.

“At the end of the day, all of these decisions are of immense public consequence and should be explained and rationalized in the public eye,” he said. “And that’s not happening.”

—Garance Burke, Associated Press

AI’s new business model: Break the internet, then sell the fix

1 week 1 day ago

People who work in cybersecurity or work on the fundamentals of digital infrastructure do not want to live in “exciting times,” but right now, they don’t have a choice.

Binance was just hacked in what is, so far, the largest cryptocurrency heist in history. Anthropic, the company behind Claude, narrowly averted a major attack through a third-party extension that would have allowed hackers to upload worms to the affected computers. A number of major airlines were affected when the Spanish travel-tech company Amadeus was hacked, leaking customer information on the web (mercifully without payment info, this time). At the center of this are generative AI companies like Anthropic and OpenAI, which all at once are empowering hackers while billing themselves as the only line of defense. They’ve found a new niche, and it’s to become a protection racket. 

The threat is quantifiable: Severe bugs are classified as CVEs (common vulnerabilities and exposures), and around 400,000 have been officially logged since 1988. Of those, 13,000 happened in the last month. PHP2Shell is the perfect example of this. If you spent any time on social media in the last few weeks, you probably heard some chatter about PHP2Shell and just how dire the threat was. Roughly 70% of websites rely on code written using PHP, and the announcement of a fundamental bug in the language—akin to discovering that millions of buildings share the same faulty lock, potentially giving intruders easy access—had people panicking. Websites such as Wikimedia preemptively went offline until a fix was pushed through.

These bugs were rediscovered separately by two cybersecurity firms: Calif and Wiz. The bugs are very difficult to spot. They had to do with how the host computer accesses memory while an application is running: If a malicious actor forced an out of memory condition by whatever means, it would be possible to interrupt the runtime, leave a partially initialized object behind, and have the application treat it as valid. It would be a stepping stone for a hostile takeover of the host server, while a separate bug could leave instructions from a request that could be read again subsequently instead of being reset. This can be an apocalyptic threat if somebody can actually do those things. They would gain remote control over the host computer, and at that point they can do whatever they want with that machine and what’s on it.

Notably, developers at the PHP Foundation knew about this bug in 2024, but couldn’t agree on a solution for it. Every proposed fix had associated costs, and in any event, nobody actually knew how to act against these bugs. The threat was completely theoretical. It sat there, which wasn’t the optimal workflow, but it also wasn’t an omega-level threat until generative AI came into the picture. Forcing the condition while also creating the takeover wasn’t easy to do. Once the vulnerability was publicly announced, what had been a trickle of attacks became a flood. Fixes were rapidly pushed (in some cases by volunteers) and anybody running a website on PHP was urged to quickly upgrade to protect themselves.

This was just one bug, but pretty much everybody expects that more will be found. The people talking loudest about the danger are, surprisingly, generative artificial intelligence companies. AI companies are all too happy to make clear how incredibly dangerous all of this is and how many more bugs and vulnerabilities they expect to find as they continue to look around. Sam Altman doesn’t think we’ll be able to patch all the bugs, while Jason Clinton of Anthropic terms it a “vulnerability tsunami.” They’re promising years of instability following this as they struggle to discover and patch all of these bugs—an ongoing process, because software development is itself a process of continual updates. 

Why are they talking this way? They’re emphasizing how dangerous their product is, which seems like terrible salesmanship. But they’re also peddling the solution to the problem they helped to introduce.

And to force the issue, they’re going to look for the bugs themselves. That’s the point of OpenAI’s Patch the Planet initiative, which is all about looking for defects and vulnerabilities in open-source software. Nor are they just running around bug-hunting and letting chaos ensue. They’re trying to step into the cybersecurity space and set themselves up as the best solution to the problem.

OpenAI unveiled a new security-focused model in September, GPT-6 Cyber. OpenAI’s Daybreak program is intended to be available only to eligible organizations so that only “good actors” can take advantage of this, though in the long term as models become open access it’s not clear how that distinction will matter. Anthropic has its own model, Claude Mythos Preview. Google has Gemini Carbon 4 as its cybersecurity model. As with all models, you pay per token. Access to Daybreak is subsidized, although if you’re a believer in Cory Doctorow’s theory of enshittification, the long-term business model will be to make this more expensive over time as customers grow to depend on it. These companies are desperate for new sources of revenue even as overall revenue growth slows.

They’re not being subtle about this. Sam Altman told a group of business students at Stanford that “I think we’re going to have to completely switch the defense model to using the same technology for a new kind of defense shield.” In other words: Gotta use my products to keep yourself safe. What they’re doing verges on extortion; they’re trying to monetize the chaos their products are introducing. But in addition to that, they’re taking aim at a fundamentally vulnerable digital space.

A threat to the open-source commons

Let’s take a step back and consider how the internet, much of which runs on open-source software, is really maintained. PHP is supported by the PHP Foundation, which has just over a dozen employees and relies on donations. It is in effect a passion project for a few developers. There’s an Xkcd comic that sheds a great deal of light on this: Random critical pieces of digital infrastructure are being maintained by people entirely on their own, thanklessly and often without compensation. The modern open internet has been built up in ad hoc ways that often deprioritized security, but that mattered less when only human beings were doing the building or the hacking. To take advantage of those vulnerabilities would require not just expertise, but a great deal of time, effort, and computational power in many cases. PHP2Shell’s exploits were hard to spot, even if you really knew what you were doing and understood PHP down to the source code.

Yes, these potential threats are dire, but this isn’t a tsunami or a natural disaster that simply could not have been prevented. This is entirely manmade, and the people shrieking about what a dire emergency it is are also the ones who have helped to inaugurate it. This exploit sat virtually unknown by anybody for more than a decade, and even when it was recognized for what it was, the threat was purely theoretical. Now because of generative AI, cybersecurity is starting to favor the attack rather than the defender. People like Sam Altman helped to create the problem, first by developing the models, and now by going around and actively looking for weak points in our ramshackle internet. They broke it, and now they’re here to offer a solution: “Nice open internet you’ve got right there; it would be a shame if something happened to it.”

And what are the costs associated with this? For individual consumers of course, there’s the threat of hacks, theft, and general instability on the internet. For businesses, outages, lost revenue, and theft. That’s severe. But I think that there’s a more fundamental risk to the whole of the internet, at least as we know it. Groups like the PHP Foundation or the individual developers who sustain free and open-source software (FOSS) are going to be under even greater strain in the years to come, especially when compared to companies like Anthropic that, for the moment at least, have very deep pockets. Those 13,000 CVEs I mentioned at the beginning of the piece? Those largely fall on volunteers and overworked organizations to remediate, which will burn them out and in the long run likely push people away from the whole FOSS project. 

Turning to generative AI is probably what Anthropic and every other company wants, as another use case for their products, but it poses its own severe risks. Large language models (LLMs) are already being used on open-source projects that call into question the whole license-free structure and system of attribution that has made FOSS possible. That’s one leg of our stool that’s being kicked out from under us.

On top of that though, as AI-generated code winds up in more FOSS projects, is it going to introduce more vulnerabilities that will then have to be hunted down and eliminated? The answer seems to be yes. Whether it’s from vibe coding or simply the fact that LLMs trained on code with security vulnerabilities will tend to reproduce those vulnerabilities, we’re likely going to be running in place for years on end. And even if it can be introduced without undermining the security of the code itself, it’s another avenue to eat away at the licensure and copyright-free status that makes FOSS possible.

So this protection racket that’s forming is more than just a tax. It’s going to massively destabilize the ecosystem of the internet as we know it. I don’t know what comes next if the delicate FOSS sandbox falls apart. Maintainers could abandon their projects, which in turn creates more security vulnerabilities as things fall out of date and become obsolete. Even stably running pieces of infrastructure will eventually be unsupported by the languages that they run on; people can wake up one morning and find that some odd dependency powering their website stopped working. Who’s responsible at that point for fixing it? 

Nor do I want to see it entrusted to for-profit entities. Maintaining this work is basically thankless as it stands; it’s unprofitable to take on because the whole point is that everybody can freely use it. The only reason for them to do so in the long run would be to monetize it, which costs the rest of us. And as they make these into their own proprietary systems, will they wall off what is theirs from everything else? Why shouldn’t they; Microsoft spent years fighting the FOSS movement because it saw open-source software as a threat to its copyrighted systems. What has been up until now a true commons, maintained by people and organizations largely doing it out of passion, would become something else entirely: less able to talk to each other, and more expensive to use.

What’s the solution? This is probably a whole other essay, but in the long run it’s going to have to be government. Whether by directly assuming the burden or offering cash and resources to private groups, governments need to take steps to ensure that an open and free internet continues to exist. Could something like the United Nations do it, ensuring that it’s a truly global movement instead of under one national or regional umbrella? Failing that, could the European Union take it on—a prudent move, especially as Europe tries to get out from under the thumb of U.S. big tech. Food for thought, because under the shadow of all of these scary cyberattacks, that’s where I think the real threat lies.

Meet the California Science Center’s new gateway to STEM careers

1 week 1 day ago

A film about the space shuttle ends with a majestic launch and rocket exhaust filling the screen. Suddenly, real-life fog envelops the theater, before the walls part and the haze dissipates to reveal a gasp-inducing 200-foot Space Shuttle Endeavour, nose up, facing skyward.

This Hollywood finish—Endeavour is the only authentic space shuttle presented with fuel tank and solid rocket boosters stacked for launch—is the centerpiece of the California Science Center’s new $450 million Samuel Oschin Air and Space Center, which opens November 13 after more than three decades in the making.

More than 200 aerospace artifacts and interactive exhibits fill three airy, multilevel galleries that illustrate the history and physics of flight, space exploration, and autonomous craft. But it’s the spectacle of Endeavour that symbolizes the CSC’s mission as Southern California’s gateway drug to science. The air and space wing comes at a time when the state’s $366 billion space technology sector is desperate for more STEM talent in a region often regarded as the epicenter of the space startup economy.

“We want to be a feeder into all kinds of science, engineering, and STEM careers,” Jeffrey Rudolph, the CSC president and CEO, tells Fast Company. “Getting particularly young people, who may not have thought of themselves as belonging in science and engineering, interested and involved. Even if they don’t want to go into science, we think it’s good to have a basic understanding of it.”

[Photo: Mike Kelley, courtesy ZGF Architects]

“There’s a war on science going on in our country right now,” the California governor, Gavin Newsom, said during a press preview last month. “That has implications not just for air and space, but for healthcare, all kinds of decisions. We want to help change that. This is a physical manifestation of what California values.”

The CSC’s mandate includes reducing initial barriers to science and guiding STEM career aspirants to their next steps. Beyond free admission, CSC offers a summer science camp, professional development for science educators, and high school internships that expose students to local aerospace companies and institutions.

California Science Center [Photo: Courtesy of California Science Center] The study of flight

The Oschin wing’s impact on science advocacy was in evidence even before its opening. In July, Astronauts for America used the shuttle as a backdrop to kick off its American college tour encouraging students to vote. Kenneth Phillips, the physicist who is CSC’s curator for aerospace science, designed the Oschin exhibits around a freshman course he teaches at the University of Southern California, even enlisting students to prototype several interactive exhibits. The opening expects some 30 astronauts, including Endeavour’s final mission crew, led by Mark Kelly, now a U.S. senator from Arizona. 

The 200,000-square-foot expansion (as big as three and a half football fields) roughly doubles the Los Angeles science center’s exhibit space, boasting a dizzying array of air and spacecraft that hang from the ceiling and jet up from lower floors. There are authentic Gemini 11 and SpaceX Dragon cargo capsules, a 3D-printed 60-foot Electron rocket from Rocket Lab, Joby eVTOL (electric vertical takeoff and landing aircraft), a drone display, and immersive experiences like an International Space Station fly-through simulation and a flight to Seoul inside a Korean Air Boeing 747-400 cockpit. A collaboration with NASA’s Jet Propulsion Laboratory and the California Institute of Technology describes their planetary exploration and Earth observation work through replicas of lunar and Martian probes, snakelike exploration robots, and the opportunity to program a Mars rover mission.

[Photo: Mike Kelley, courtesy ZGF Architects]

The 20-story Endeavour, alongside its orange propellant container and two solid rocket boosters, holds court in the cavernous shuttle gallery. Its base lies 25 feet below, accessible in part by a dark 45-foot tunnel slide with turns that replicate the shuttle’s reentry path, an idea borrowed from Kennedy Space Center’s Atlantis shuttle showcase. (“We saw that and said, ‘That’s good, but we can make it bigger and better,'” Rudolph laughs.) Several levels offer close-up views into the cockpit and mid-deck, and a payload bay stocked with supplies; above, there’s a glass platform from which visitors can look down on the entire shuttle.

“We wanted to show the whole system in launch position to explain the science and engineering behind it and give a better understanding of the enormous amount of energy it takes to get off the surface of Earth and into orbit,” says Rudolph. “And it is the most majestic way to see it.”

[Photo: Mike Kelley]

These views, once reserved only for NASA engineers, astronauts, and their families, “we now get to share with the rest of the world,” says the retired NASA astronaut John “Danny” Olivas, who flew two shuttle missions and advised on the Oschin Center. “The visual experience does a much better job of capturing the enormity of it.”

We got the space shuttle. Now what?

Getting hold of a retired shuttle to display in launch position was always part of CSC’s master plan, adopted in 1993, and the center won Endeavour from NASA in 2011 in a competition among museums. (The other three shuttles—Atlantis, Enterprise, and Discovery—reside on the East Coast.) Aside from the state’s long history with the aerospace industry, the shuttles were designed, built, and often landed in California.

“So, we got it,” says Rudolph. “And then we’re like, ‘Now what?’”

CSC brass set about raising money, figuring out transportation logistics, locating or fabricating missing shuttle pieces, and designing the first building to house a vertical shuttle outside of a NASA or Air Force facility, feats requiring the guidance of NASA engineers. Funding came from its namesake, Samuel Oschin, the Los Angeles entrepreneur who died in 2003; the state of California; state-backed bonds; and private donations, which are still being raised. (Also, some exhibit elements come from commercial and institutional partnerships.) Meanwhile, they had to get Endeavour to Los Angeles, and to the center.

“We had no idea how complex it was. We probably wouldn’t have done it if we did,” says Rudolph. “A lot of the folks at NASA didn’t think we could pull off getting it from LAX to the science center, because you can’t take it apart like an airplane. We’d never be able to put it back together again.”

NASA’s Shuttle Carrier Aircraft with space shuttle Endeavour atop. [Photo: NASA]

In 2012, the shuttle—ferried atop a modified Boeing 747 shuttle carrier aircraft—famously flew around California, over NASA facilities and iconic landmarks, before landing at Los Angeles International Airport. Then it was transported, on a specialized rolling steel platform, slowly through a 12-mile nonlinear street route that had to be more than 78 feet wide with no overhangs below 60 feet high, passing 1.5 million cheering onlookers. The tank came by barge through the Panama Canal, up the West Coast to Marina Del Rey, before moving through the streets. “Almost half a million people showed up—for a fuel tank,” says Rudolph.

After more than a decade in a horizontal stance in a CSC atrium, Endeavour was ready for its final liftoff. In early 2024, a massive 450-foot crane hoisted the shuttle by sling, turned it upright, and slowly lowered it—with an inch to spare on either side—onto an isolated concrete seismic slab on ball bearings to move with earthquakes. Given the beating it had to take from launch and reentry, “we figured that the shuttle as a design can withstand the worst earthquake anyone’s predicting for L.A.,” says Dennis Jenkins, the Oschin Center project director, who spent three decades as a NASA contractor working on the shuttles.

The Oschin Air and Space Center’s long journey to realization ends with a first step for a new generation of explorers, enthusiasts, and even creatives seeking inspiration from new frontiers. Eventually, the air and space center hopes to collaborate with the nearby Lucas Museum of Narrative Art, Natural History Museum of L.A. County, and California African American Museum. 

“The country really needs science and engineering more than ever now, but it’s more than just STEM,” says Rudolph. “Space inspires across the entire spectrum of human expression.”

Your city’s bike network is broken. These tools could help fix it

1 week 1 day ago

In most cities, bike networks are rolled out in disconnected pieces. A bike lane on one street might not reach the next part of a route, so you can’t necessarily ride safely across town. A new set of software tools helps planners, bike advocates, and citizens design more complete networks, automatically filling the gaps in existing infrastructure or creating the whole thing from scratch.

Called BikeNetKit, the free, open-source software is based on eight years of research on bicycle infrastructure at Denmark’s IT University of Copenhagen.

When you type in a city name, the software pulls street data from OpenStreetMap, including a map of existing bike lanes. (Getting started requires a little technical knowledge: You have to know how to use Python, and then have to enter a few lines to install and run the algorithms.)

[Image: IT University of Copenhagen]

“Planners mostly think about the small picture—what if I add a lane here? What’s on the right side and left side? But they don’t think about the network and the whole city,” says Michael Szell, an associate professor at the IT University of Copenhagen, who led the project. “Our methods are built on network science methods that look at the city from a big picture.”

[Image: IT University of Copenhagen]

One of the tools, called GrowBikeNet, can build a network from scratch—particularly useful in the U.S., where most cities have sparse bike lanes. The algorithm creates a grid of “seed points” over a city. It connects those points to any existing bike infrastructure, drawing a new map of suggested bike routes.

[Image: IT University of Copenhagen]

“The first links that are suggested are the links where you expect the most cyclists to ride,” Szell says. Rolling it out this way would create a functional network from the outset, with strong connections already in place.

[Image: IT University of Copenhagen]

Users can upload additional data to help plan a network—for example, where traffic crashes have occurred in the past, or where people typically travel each day. The software also gives the option to plan bike lanes around schools. Another tool, called Superblockify, suggests places to add car-free neighborhoods; LinkBikeNet connects gaps in otherwise well-connected networks, while FixBikeNet tells cities which gaps are most important to fix first.

[Image: IT University of Copenhagen]

The tools don’t specify what the bike infrastructure should look like. Instead of bike lanes at the side of the road, for example, some streets could be dedicated solely to people on bikes and pedestrians. Of course, that’s the opposite of most transportation planning now, which focuses on the flow of cars. “Planning for cycling is always some kind of afterthought,” says Szell.

The BikeNetKit website includes hundreds of analyses of European cities, from growing a new bike network in Kyiv, Ukraine, to fixing gaps in Copenhagen’s network, to potential car-free neighborhoods in Berlin. To try it out yourself, the open-source code is available on GitHub.

3 calendar and text scams circulating right now, and how to avoid them

1 week 1 day ago

We’ve all gotten the text message claiming a package couldn’t be delivered because of an incomplete address. But scammers are getting a lot more creative, and they’re weaponizing the built-in notification systems on your phone to get your attention.

Instead of just sending emails that end up in your spam folder, they’re injecting fake events directly into your calendar and spoofing group text messages to bypass your carrier’s basic filters.

Here’s how to spot the newest wave of mobile scams and quarantine them without compromising your device.

Nuke rogue calendar subscriptions

If your calendar is suddenly flooded with alerts claiming your phone has a virus or you’ve won a prize, you haven’t actually been hacked. You accidentally clicked a malicious pop-up on a website that subscribed you to a spam calendar.

To fix it on an iPhone, go to Settings, tap Calendar, and tap Accounts. Look for Subscribed Calendars, find the one you don’t recognize, and tap Delete Account. The spam will vanish instantly.

On Google Calendar, go to calendar.google.com on a computer (it’s much easier to manage subscriptions on a computer than the mobile app). Look at the left sidebar under Other calendars. Hover over the spam calendar you don’t recognize and click the X icon to remove the calendar.

Ignore the address update text

The package delivery scam is incredibly effective right now. You’ll get a text claiming a delivery is suspended, complete with a link to a fake postal service website asking for a small redelivery fee.

The postal service doesn’t send texts from random 10-digit numbers. Never click the link. Instead, copy the tracking number if there is one and paste it directly into the official shipping website yourself to see if it’s legit or not.

Silence the chaotic group text spam

Scammers are now dumping 20 or so random phone numbers into a single group text containing a malicious link. Once one person replies to ask who everyone is, your phone starts buzzing nonstop.

Don’t reply. Open the message thread, tap the group icon at the top, scroll down, and tap Leave this Conversation or Leave Group. Reporting the group as spam produces a similar result.

Enterprise AI is waiting for its Transformer moment

1 week 2 days ago

As a professor of innovation, I have a slide in my deck I use every time I start a session on artificial intelligence: a fragment of the first page from the “Attention Is All You Need” paper, from 2017. I use just a fragment because I don’t want the students to read it. . . . I usually refrain from using a lot of academic papers when I teach MBAs and executives; they tend to be boring, and this one is no exception and is not particularly easy to read. However, it is historic. Written by eight Google researchers, it is the paper that defines the Transformer architecture, which among other things, became the architecture behind GPT and much of the subsequent generative AI wave. The rest, as I said, is history.

The Transformer was not only a successful model architecture; it was also a huge architectural discontinuity. The 2017 breakthrough was able to completely reorganize computation around self-attention, eliminate recurrence, and make parallel training much easier. And in doing so, it did not simply enlarge what was already there; it completely changed the architecture. 

In 2017, Google understood the technology, but underestimated the category it could create. It did not ignore the Transformer; it incorporated the architecture into products such as Translate and Search, but it failed to turn its own breakthrough into the product that defined the next era. OpenAI, founded two years before the paper appeared, recognized its broader potential and built GPT, and ultimately ChatGPT, on top of it. Meanwhile, all eight authors eventually left Google, seven to become founders and one to join OpenAI. For present purposes, “Transformer moment” means an organizing principle whose eventual importance greatly exceeds its initial application. 

Making an artificial worker more capable vs. making an organization intelligent

I have been insisting on this for a long while now: An individual, human or artificial, can complete tasks, but a company, on top of that, must preserve state, coordinate specialized capabilities, enforce authority, remember outcomes and adapt across time. In that sense, transformers provided the architectural foundation for extraordinary advances in model-level and task-level capabilities. But from there, there is still a significant gap before you reach enterprise performance.

In McKinsey’s 2026 survey, 80% of respondents reported improved personal productivity, but only 37% reported a positive contribution to enterprise EBIT [earnings before interest and tax], with no significant differences from the previous year. This reveals the distance between individual augmentation and organizational results. McKinsey also notes that the high performers are typically those companies that redesign workflows, rather than simply inserting AI into existing ones. 

In practice, the industry’s current starting point seems to be the agent. Anthropic explicitly begins with the “augmented LLM,” a model supplied with retrieval, tools and memory, while Microsoft presents sequential, concurrent, group-chat, handoff and other multiagent patterns. These are useful techniques, but nearly all begin with an artificial worker and ask how to connect it to other workers. Is coordinating agents the same thing as designing an intelligent organization? To me, at least, it looks completely different. 

This proliferation of patterns is the evidence of an unsettled foundation

The problem is not just variety. What fails is that every implementation, each agentic project, repeatedly reconstructs the same underlying capabilities, from identity, to permissions, state, policy, evidence, recovery and learning, because no common foundation exists. 

If we take a look at Gartner’s forecasts, they predict that more than 40% of agentic projects will be canceled by the end of 2027 due to cost, unclear value, or inadequate risk controls. This supports the idea that model capability alone is insufficient. Gartner also explicitly recommends focusing on enterprise productivity as a whole, rather than individual task augmentation.

What’s the real problem, then? Simply put, that the agent should not be the organizing unit. The center should instead be a persistent, goal-directed organizational system. Models and agents can contribute reasoning or execution, but the company-level system owns the representation of the business, its operational state, objectives, constraints, authority and accumulated experience. 

In order to work properly for an organization, such an architecture should have persistent organizational state; formal objects, relationships, and permissible actions; objectives and trade-offs; deterministic authorization outside the model; ephemeral agents and credentials; escalation mechanisms; and an append-only record connecting intent, action, and outcome. Recent guidance from NIST [National Institute of Standards and Technology] strongly supports architectural separation between reasoning and policy, deterministic access decisions, and immutable evidence. Microsoft’s reference architecture similarly separates the agent runtime, policy layer, tools, systems of record and audit pipeline. 

What should be the test for a genuine architectural breakthrough?

The evidence to prove a breakthrough should therefore not be whether a system outperforms an agent on some benchmark. A genuine breakthrough should make several apparently separate problems — coordination, governance, memory, replaceability, accountability, and continuous learning — much simpler to manage, because they are handled by the same underlying design.

Google invented the Transformer and successfully deployed descendants such as BERT, but it failed to capture the defining consumer product moment in the way OpenAI did, perhaps because they were uncertain about the effects of the technology on their business model, like what Clayton Christensen exposed in The Innovator’s Dilemma. Large language models were much more expensive to run for web searches, provided a complete answer in such a way that most users wouldn’t bother to click on an ad, and sometimes hallucinated, endangering the confidence in their results that Google had spent decades building.

Later, Alphabet’s hurried response with the Bard chatbot culminated in errors and roughly $100 billion in lost market value in February 2023. As we can see, research leadership does not automatically translate into product or category leadership. Google has improved a lot since then and has outperformed other competitors on several occasions, but at the time, they lost an opportunity to exploit what could have been a clear pioneering advantage. 

When Larry Page and Sergey Brin presented PageRank [in the 1990s], they tried to license their technology to several companies, the internet behemoths of the moment. Yahoo! declined even to meet; Excite doubted PageRank’s business value and declined to license it. Those incumbents evaluated an unfamiliar technological advantage as if it were another search technology whose value they doubted, and continued with what they knew worked for them . . . until it stopped working. Again, the rest is history. 

Will today’s AI leaders recognize organizational intelligence as a new category? Or dismiss it because its first implementations appear narrower than the eventual architecture? By insisting on putting the LLM in the center of everything, they are making the same mistake that these internet behemoths committed with Google, and a similar one, misreading the category, that Google committed with the Transformer. Placing the LLM or the agent at the center risks optimizing the previous paradigm instead of discovering the next one. 

The current agent-centric paradigm leaves a clearly identifiable architectural gap. Whoever supplies the organizing principle for intelligence at organizational scale will have the opportunity to completely reshape the industry. And the missing architecture will not necessarily be built by today’s model leader. 

The only question is whether today’s AI leaders will recognize that moment, or provide another company with the opportunity to define it for them. 

5 books to calm AI-induced existential dread

1 week 2 days ago

Artificial intelligence is transforming how we work, connect, and live. But with its extraordinary potential come some big questions about trust, human relationships, and the future of society. These five books explore the promises and perils of AI, offering fresh perspectives on a technology that’s changing our world faster than ever.

Learn something new every day with “Book Bites,” 15-minute audio summaries of the latest and greatest nonfiction. Get started by downloading the Next Big Idea app today!

Artificial Intimacy: Who We Become When We Talk to Machines

By Sherry Turkle

Tracing our new habit of talking to machines through the human life cycle―from children’s earliest attachments to how we face death―Turkle offers both a cautionary tale and a road map for reclaiming our humanity in the age of AI. Listen to our Book Bite summary, read by author Sherry Turkle, or view on Amazon.

Trias Algorithmica: What Code Rules

By Hamilton Mann

Whether you oversee digital transformation, shape regulation, or design algorithmic tools, this book provides the conceptual and practical foundations for accountable algorithmic governance. Listen to our Book Bite summary, read by author Hamilton Mann, or view on Amazon.

The Eureka Machine: Why AI Is the Key to Unlocking a New Era of Scientific Discoveries

By Richard Socher

The most meaningful application of AI will be the automation of scientific discovery itself: a Eureka machine that combines all human knowledge, simulates billions of experiments, and generates breakthroughs. Listen to our Book Bite summary, read by author Richard Socher, or view on Amazon.

Manage the Machine: How to Harness Human-AI Collaboration at Work

By Paula Goldman

From Salesforce’s chief ethical and humane use officer, a prescriptive guide for how to integrate AI into the workplace effectively. Listen to our Book Bite summary, read by author Paula Goldman, or view on Amazon.

Human Raised: Nurturing Connection, Curiosity & Lifelong Learning in the Age of AI

By Dana Suskind

The challenge isn’t to reject AI as a parenting tool—it’s to use it in ways that strengthen, rather than substitute for, the messy, imperfect interactions children need to flourish. Listen to our Book Bite summary, read by author Dana Suskind, or view on Amazon.

This article originally appeared in Next Big Idea Club magazine and is reprinted with permission.

Enjoy our full library of Book Bites—read by the authors!—in the Next Big Idea app.

7 Mac and iPhone apps worth downloading right now

1 week 2 days ago

This article is republished with permission from Wonder Tools, a newsletter that helps you discover the most useful sites and apps.

I find, try, and build apps nearly every day for my Hulry newsletter. Here are seven iOS and macOS apps I recommend, including one I made for my own weekly planning.

Dia: a clean browser with optional AI

Brave, my previous browser, started glitching and lagging, so I switched to Dia.

When I first tried Dia in 2025, I wasn’t a fan because it was AI-first. I didn’t want that. It has since become a well-made everyday browser with optional AI features. You can get a morning briefing from your connected apps, or ask questions about the page you’re reading.

It’s clean, looks great, and is regularly updated with tiny, useful features. For example, I love how it automatically groups the tabs I open from the current page with Command-click, then gives the group a name.

Platform and Price: macOS | free; $10/month plan optional

CleanShot: screenshots and recordings with smooth zooms

I’ve used CleanShot as my screenshot tool for more than seven years. I use it for blog posts and tutorials, and to capture payment confirmations in case of later disputes.

CleanShot could already record my screen. Its new Studio Mode is a big update for video editing, including smooth zoom effects that make recordings look more polished.

I use these zoom-in effects in my tutorials and in my social media posts to draw viewers’ attention to the important bits of the recording. I used to use Screen Studio for that editing, but I’m gradually shifting to using CleanShot for everything.

Platform and Price: macOS | $35 onetime purchase, including a year of updates

PoseKit: turn iPhone captures into polished demos

While CleanShot is fantastic for making screen recordings on a Mac, PoseKit fills the gap on an iPhone.

I’ve been using this app to make good-looking demo videos and app mock-ups for social media. It consistently works well.

I like how easy it is to add angled views, backgrounds, and gradient blur to iPhone screenshots and screen recordings.

While it’s most useful if you have an app to showcase, it can also be handy for making tutorials for YouTube or other social platforms. You can record your iPhone screen as usual, then quickly wrap a sleek iPhone frame around the video.

Platform and Price: iOS and iPadOS | free download; $13 Pro upgrade

Atlas: collect visual inspiration on a board

If you collect visual references for work or hobby projects, try Atlas. It’s one of the most polished apps I’ve used.

It’s easy to save new clippings, and the interactions are remarkably smooth and fast. I love the view that lays out your saved items on an infinitely scrollable digital whiteboard. That gives you a bird’s-eye view of everything you’ve collected for a project.

It’s ideal for home renovations or even craft projects. Collect example designs as you browse the web or social media. Then, once you have a sizable collection, lay everything on an infinite board for design inspiration.

Platform and Price: macOS | $39 onetime purchase, including a year of updates

Spatial: organize notes and links on a canvas

Atlas is primarily designed for organizing visual references, while Spatial is excellent for other materials, like notes, color palettes, and links.

You get an infinite digital canvas where you can save sticky notes, images, videos, and documents, then organize them into folders.

Platform and Price: macOS | Pro: $20/year or $50 lifetime

Refine: a Grammarly alternative that runs on your Mac

I’ve long been looking for a Grammarly alternative that runs on my own computer, not online, so my writing stays private. None came close to Grammarly’s results, not even the open-source checker Harper.

Refine is the alternative I was looking for. I use it with Google’s free Gemma 4 AI model on my M1 Pro MacBook. It has replaced Grammarly for me.

Platform and Price: macOS | free; $38 lifetime upgrade

Foreweek: a paper-style planner for weekly goals

Disclosure: I made this app.

I set tiny goals at the start of each week based on what would make my week feel successful and well spent.

I used to plan on paper. Foreweek re-creates that feeling. Each week, you get a blank page to plan your goals, not tasks. If you have unfinished goals from previous weeks, you can add them to your current list.

Platform and Price: iOS | $2.99 onetime purchase

This article is republished with permission from Wonder Tools, a newsletter that helps you discover the most useful sites and apps.

5 tips for getting the most out of AI

1 week 3 days ago

Why do some people seem to be getting a lot out of AI, while others—maybe you—still feel like they’re spinning their wheels trying to squeeze anything remotely useful out of these platforms? The answer is multifaceted, of course.

Some workflows are better suited to the strengths of large language models, while others are more vulnerable to their weaknesses. Some users pay for upgraded plans, which can improve how well the models work.

But a big part of the difference comes down to how you use artificial intelligence. Like any tool, there are ways to make it perform better for you. In other words, the technology is often what you make of it.

If you offload your thinking to AI platforms, you risk dulling your ability to approach problems creatively and becoming dependent on a flawed technological crutch. But if you use AI to challenge yourself, stress-test your work, and outsource routine tasks, you might feel, at least in some sense, like you’ve gotten a significant upgrade.

Consider the calculator: Some people have become so reliant on it that their mental math skills suffer, while others use the technology to tackle more advanced math because they’re not spending their time on simpler problems.

Like all new technologies, what you choose to outsource to AI is up to you. Still, if you do decide to use it, it’s worth considering how to get the most out of the technology—and make sure chatbots serve you, rather than the other way around. Here are a few tips to make these platforms work better for you.

Be smart about your prompts

What you type into your AI actually matters. Take traditional search engines. You may not realize it, but over time, you might have developed an implicit sense of how using certain Boolean operators or more specific proper nouns can yield better, more relevant results to your queries.

The same is true for AI. Being more specific with a chatbot about what information you want can make a big difference, particularly when you’re using the technology as a search tool. It might be easy to slip into thinking that AI should already know what you want, but it doesn’t. Instead, treat a chatbot as an earnest research assistant with the capacity to make a lot of mistakes. There’s an old adage: “The more you know, the more they know.” The same logic applies to AI.

Here’s an example: If I’m shopping for a bookshelf, I no longer sift through countless listings searching for something that feels right. Instead, I take a picture of the spot where I need it to go, give the AI the dimensions (or range of dimensions) I’m looking for, and ask the chatbot to find the best options given those constraints. This saves me a lot of time and allows me to work smarter when using AI for shopping.

Encourage humility

One of the big problems with AI—or any work you don’t do yourself and instead outsource to someone (or something) else—is that it will make mistakes. Fear of an AI oopsie is anxiety-inducing and absolutely real, and one of the best reasons not to completely offload all your tasks to ChatGPT or Claude. Really, don’t do this!

The best approach is to check any work produced by a chatbot yourself. But there are also ways to help these tools work better. If you have the ability to adjust your AI’s settings, tell it to be more humble—and more cautious—when sharing information or making claims. Tell your AI to consider multiple ways of thinking about a question or problem, and to present divergent viewpoints. For kicks, tell it to admit when it’s unsure.

Also, always ask for direct citations and check them off the platform. When you send a prompt to your chatbot, tell it to avoid simply summarizing information and instead to provide direct links to sources so you can verify the claims yourself. Once you do this, you might be shocked by how often AI claims to be consulting a source or generates a fake web page only to apologize when you discover it doesn’t actually exist.

A polite reminder: Just because Claude says it’s “double-checking” something doesn’t mean it actually is. You need to double-check it, too! AI can cite unreliable sources and even hallucinate citations entirely. AI is powerful, but also unpredictable, and should never be treated as an objective source of information. Think of it as a sieve that helps you sift through information and find what you want faster, rather than something you can trust to identify the best results on its own.

Understand memory

Your AI is remembering what you ask it and what you tell it, even if you’re not always aware. This isn’t necessarily unique to AI systems; many tech platforms collect all sorts of data about you and then use that information to tailor content to you (usually, ads). It’s possible you don’t want AI to remember anything about you at all, in which case, check your privacy settings—or perhaps avoid AI entirely.

But if you are comfortable with a chatbot retaining information about you, consider that its memory can actually be a useful tool. For instance, if you’re using AI to draft emails, tell your chatbot to always introduce you in the way you prefer. If you tend to offer to jump on a call or Zoom, tell the AI that you like doing so—and ask it to remember that preference for when you draft emails in the future. If you prefer a more casual tone, tell it that. If you prefer to remain formal, tell it that instead.

Here’s another example: If you’d like your AI to avoid sending you links from Amazon when you’re online shopping—perhaps you don’t have an Amazon account—tell it to remember that preference so you don’t have to keep sifting through those links. Or maybe you’d prefer your AI to prioritize links to The New York Times because you already have a subscription. Tell your AI that, too.

Also, remember that you don’t need to give AI your personal information at all. For kicks, try asking it to erase things from its memory that you don’t want it to retain. There’s absolutely no guarantee that a request alone will erase the information completely, so check your privacy and memory settings, too. But why not give it a stab?

Customize how AI communicates

You spend a lot of time on Google, but you might now be pivoting to AI to surface information instead. One advantage of this is that you can have AI produce results in all sorts of formats and styles, since you dictate the terms in which it expresses itself. For instance, you might prefer to receive information in simple bullet points rather than a long, essay-like string of text. You can share that preference with your chatbot.

Maybe you prefer visuals and graphics? AI is great at whipping those up. For the nerds: Maybe you prefer receiving information in code that you can then copy and paste into a development environment, like Python, so you can run a script to check whatever work the AI did in a more deterministic way.

There are even simpler changes you could make. You might prefer that AI respond in two languages, perhaps because you already know one but are working on developing your skills in the other. Seeing the text side by side as you interact with the platform throughout the day might help boost your fluency.

Another example: I’ve instructed my AI to regularly incorporate more unusual, niche vocabulary. There’s a somewhat well-founded fear that AI is homogenizing our language and making us less creative in how we express ourselves. I told AI to reverse that trend and gave it a list of common AI crutch words to avoid.

Know your way around tokens

AI operates on tokens. You spend input tokens on the information you give a model, including your prompts and relevant context, and output tokens on what it produces in response. You may not always know how many tokens you’ve used, but writing long prompts generally means using more input tokens, while receiving long responses means using more output tokens.

As you employ chatbots and run up against usage limits, consider that you can guide your AI on how much detail to provide for a particular task. Tell it to be brief when you want a short response or just one link, and tell it when you want a really detailed answer.

Beware this CAPTCHA scam designed to steal your data

1 week 3 days ago

CAPTCHAs are one of the great annoyances of the modern internet, yet they play an important role in keeping the web secure. These tests, which usually involve solving simple puzzles or deciphering distorted text, help websites determine whether a visitor is human, thus helping prevent malicious bot attacks.

But now scammers are using the ubiquitous test to attack the humans—and if you fall victim to a CAPTCHA scam, all your data could be up for grabs.

What is a CAPTCHA scam?

To understand what a CAPTCHA scam is, you first need to understand what a CAPTCHA is. You’ve certainly seen one even if you don’t know what it was called. A CAPTCHA is that little box that sometimes appears on a web page that asks you to type out text it has purposefully distorted or click images that are all alike in someway—such as ones that feature buses or bridges—displayed in a 3×3 grid.

CAPTCHAs—acronym for “Completely Automated Public Turing test to tell Computers and Humans Apart”—help website owners ensure that a human, and not a bot, is accessing the site’s information. This helps protect against automated scraping and denial-of-service attacks.

The ubiquity of CAPTCHAs has conditioned people to expect them—and that’s enticing to scammers, who now use fake CAPTCHAs to get unsuspecting users to install malicious code on their Macs and PCs, designed to steal their data. Fake CAPTCHAs are becoming so prevalent that the Federal Trade Commission (FTC) released an alert earlier this year warning consumers of the tactic.

How does it work?

The CAPTCHA scam shows a fake CAPTCHA pop-up on a malicious website that looks like the real website for a legitimate company, such as Amazon.com or Facebook.com. But instead of asking you to enter the distorted text you see on screen or solve a simple image puzzle, the fake CAPTCHA tells you to press certain keys on your keyboard, ostensibly to help verify that you are human or to verify your account.

These keystrokes are known as keyboard shortcuts, and by entering them, you copy and paste code that the fake CAPTCHA has added to your clipboard, allowing scammers to automatically download and install malicious software on your phone or laptop.

If you’re on a Windows machine, you may be asked to press Windows+R, then Ctrl+V, on your keyboard. On a Mac, you may be asked to press Command-Space, type in “Terminal,” and then press Command+V. Windows+R opens Windows’ Run launcher. Terminal is the Mac’s equivalent. Both let users enter command-line instructions that can tell the computer to download and install files and apps.

Far from verifying that you are human, the pasted text installs keyloggers or spyware on your device designed to steal your usernames, passwords, and other data, which the scammers then often use to commit financial and other crimes.

How do I know if I am the target of a CAPTCHA scam?

The best way you can know if you are a target of a CAPTCHA scam is by being aware that the scam exists, and by understanding how CAPTCHAs actually work.

Legitimate CAPTCHAs will never ask you to carry out keyboard commands or open any type of app or command line interface on your computer. The second you see a CAPTCHA that asks you to do so—or has a long list of instructions—beware. It is a telltale sign that the CAPTCHA is a scam.

Other signs that the CAPTCHA could be a scam? When it pops up unexpectedly and claims to be for a site, like Facebook, that you are not currently on, or appears on a website that clearly isn’t the site it purports to be (like a fake Amazon.com or Apple.com website).

How can I protect myself?

Knowledge is your best defense. As soon as you think you may be being targeted by a CAPTCHA scam, do not interact with the CAPTCHA displayed on your screen.

If you have interacted with a fake CAPTCHA—and especially if you notice that doing so has caused something to download to your computer—the FTC advises you to quickly disconnect your computer from the internet (i.e., shut off its Wi-Fi connection) and immediately run security software to help find and remove any malware.

Checked
20 minutes 2 seconds ago
Fast Company Technology
Fast Company inspires a new breed of innovative and creative thought leaders who are actively inventing the future of business.
Subscribe to Fast Company Technology feed