Janko Roettgers | The Verge

BitTorrent’s disastrous, legendary, and controversial story

2026-07-02T11:42:15-04:00

Twenty-five years ago today, a young, little-known programmer by the name of Bram Cohen fired off a short message to a mailing list for peer-to-peer enthusiasts. “My new app, BitTorrent, is now in working order, check it out here,” Cohen wrote, followed by a link to his personal website.

“What’s BitTorrent, Bram?” the founder of the list asked in response.

Cohen never bothered to reply. The world would find out soon enough.

In the following years, BitTorrent quickly became the world’s most popular file-sharing app, unleashing a massive wave of piracy that upended Hollywood forever. At one point, BitTorrent was said to be responsible for a huge amount of internet traffic — some widely cited metrics peg it at half of P2P and one-third of all internet traffic in 2004. And while the entertainment industry succeeded in shutting down file-sharing systems like Napster and Kazaa, it largely failed to curtail the massive flood of BitTorrent piracy.

What stymied movie studios and record labels alike is exactly what makes the story of BitTorrent’s 25th anniversary a gripping one. There’s BitTorrent, the app that Cohen unveiled in July 2001 and which continues to attract tens of millions of monthly users to this day. There’s BitTorrent, the file-sharing protocol, which has been adopted and advanced by hackers and developers from around the world, and which helped birth an entire cottage industry of piracy websites. And there’s BitTorrent, the company cofounded by Cohen, which struggled for years to make money with its technology.

“Going into it, my plan was not to start a business,” Cohen recalls. “My plan was to start a revolution.”

This is the story of that revolution, and all that remains of it today.

Cohen started working on BitTorrent soon after leaving Mojo Nation, a startup that had ambitious plans to combine file sharing, distributed computing, and micropayments, only to shut down in 2002 without ever really getting off the ground.

Cohen helped Mojo Nation improve the efficiency of its file-sharing technology. His approach was to use something called swarming distribution: Instead of swapping large files between a few users at a time, the app would divide each file into lots of little chunks, and then allow large numbers (or swarms) of users to trade those chunks among themselves. The ethos of sharing was built into the software, insisting that users would help upload files instead of just downloading them selfishly.

Cohen was disillusioned by the failure of Mojo Nation but not ready to give up on the underlying technology just yet. “I decided to make a tool that very narrowly just did swarming distribution and nothing else, because that was a thing I knew how to do,” he says.

“My plan was not to start a business. My plan was to start a revolution.”

All this happened as P2P and file-sharing companies were facing massive headwinds. The music industry sued Napster at the end of 1999 and ultimately forced it to shut down around the time Cohen first released BitTorrent. A few months after that, the entertainment industry sued Kazaa and Grokster, followed by legal claims against Audiogalaxy, LimeWire, eDonkey, and other file-sharing systems.

Despite some significant differences in their P2P architecture, all these apps ultimately lost in court or folded ahead of a likely legal defeat. Their Achilles’ heel: They offered users a way to search for copyrighted files and then helped them find others to trade those files with. Courts deemed this contributory infringement, pointing to the fact that the owners of these services knew exactly which files their users were trading.

BitTorrent was different. From the start, Cohen built the app without search functionality, instead relying on third-party websites for content discovery. He also didn’t run a centralized server to help users discover other BitTorrent clients to trade content with, and instead outsourced that piece of the puzzle to so-called tracker servers. Websites dedicated to BitTorrent file sharing would offer a metadata file (the torrent file) for every download, which in turn pointed users to tracker servers that matched up users to facilitate the sharing. Simply put, BitTorrent was an app to help people trade data, but its developer had no way of knowing what that data was.

That structure effectively shielded Cohen and the company he would eventually found from legal liability. To this day, BitTorrent remains the only major publisher of file-sharing software that hasn’t gotten sued by Hollywood and the major labels. However, Cohen insists that the decision to outsource content search and client discovery wasn’t primarily a shrewd legal maneuver, but a simple necessity.

After all, he was just one guy. “My hands were somewhat tied,” he says. “I was writing stuff on my own.”

Despite lacking the resources of much bigger P2P companies, Cohen’s BitTorrent app quickly gained traction in tight-knit communities looking to trade large files. Among the first adopters was Etree, an online community of jam band enthusiasts. Bands like Phish and the Grateful Dead had long encouraged their fans to trade bootleg recordings of their shows with each other in the offline world. Etree brought this idea online with the help of BitTorrent, which allowed fans to share show recordings using high-resolution audio files that were simply too big to distribute any other way.

But pirated videos were like rocket fuel. “When it started getting used for anime was when the numbers took off,” Cohen says. Granted, that early success didn’t quite measure up to the tens of millions of users apps like Kazaa were attracting at the time. “It was like, Oh, look, we’re getting a thousand downloads a day, that’s massive,” he recalls.

Things changed significantly with the emergence of major torrent sites, which allowed anyone to quickly find and download movies and TV shows. In 2002, a Slovenian high school student launched Suprnova.org, which reportedly attracted up to a million users a day before its closure in 2004. And in 2003, a ragtag group of hackers founded The Pirate Bay in Sweden. With its open embrace of illicit content and an F-you attitude toward Hollywood, the site turned into a kind of lightning rod for self-described pirates and copyright enforcers alike.

“It was so fast that we initially thought something was broken.”

While these sites quickly became household names, countless others sprung up all across the globe in their shadows. Some catered to specific interests like anime, porn, or ’70s B movies, while others simply welcomed everyone and everything. And while copyright holders were able to shutter some sites, others survived, or reemerged under different names or domains. Plus, all those lawsuits generated countless headlines, effectively promoting BitTorrent to new audiences. By 2004, BitTorrent traffic exploded.

The surge of interest didn’t just alarm rightsholders. It also put some significant strain on BitTorrent’s infrastructure, including the tracker servers that helped users discover each other. One popular tracker, Denis.Stalker.H3Q.com, was run by a trio of German hackers associated with the local Chaos Computer Club.

Initially, their tracker ran server software Cohen had written in conjunction with the BitTorrent client app. But as BitTorrent gained in popularity, their server, which was just a single computer humming along in a kitchen somewhere in Berlin, was bombarded with hundreds of requests per second, slowing it to a crawl.

That’s when Dirk Engling, a member of the trio who is also known as Erdgeist in German hacker circles, decided to build a better-performing tracker software from scratch. He dubbed his server software Opentracker, and the difference was like night and day: While the Denis.Stalker server previously always ran hot, Engling’s code put a sudden stop to the near-constant fan noise. “It was so fast that we initially thought something was broken,” he says.

Engling is not as well known as Cohen. But back then, he made a fateful decision that ultimately helped BitTorrent scale to surpass all other file-sharing networks in popularity. Instead of just using his code on the Denis.Stalker tracker, Engling published it online, free to download for anyone.

Opentracker quickly took off and was adopted by major torrent sites like The Pirate Bay as their tracker solution of choice. Many of these big sites relied on third-party tracker servers like Denis.Stalker, but for years also ran their own trackers as fallback solutions. Every torrent file can point to multiple tracker servers, so sharing continued when one server went down. (Eventually, BitTorrent added further decentralization to allow trackerless trading, but tracker servers remain to this day an important part of the ecosystem.)

“Back then, it was a radical political idea,” Engling says. The thought was to allow anyone to distribute large amounts of data without having to rely on expensive servers — and if some of that was used for piracy, so be it. “It was a movement to democratize [file sharing],” he says.

Engling, for one, wasn’t in it for the money. He distributed Opentracker as “beerware,” an open-source license guided by the notion that anyone who found it useful should buy him a beer, as he half-jokingly explained in a 2007 talk. “I thought: A lot of people are using BitTorrent, so…”

Back in the United States, Cohen quickly realized that relying on the goodwill of others — especially people who used his software to download content for free — didn’t pay the bills. In 2004, he founded BitTorrent Inc. to transform the revolution he had started into an actual business. “That didn’t work so well, unfortunately,” he now concedes.

BitTorrent raised $8.75 million in funding in 2005, followed by another $20 million round in 2006. The money helped the company to staff up, but it also put pressure on Cohen and his teammates to somehow turn an app that was beloved by pirates into a legitimate business.

As part of these efforts, the company struck deals with Hollywood to launch a movie download store called the BitTorrent Entertainment Network, which launched with thousands of titles from Warner Bros., Paramount, and MGM in early 2007. However, the company quickly realized that selling DRM-protected movies to an audience that already had access to the same titles for free was a fool’s errand.

“Back then, it was a radical political idea.”

“That didn’t work at all,” says a former senior BitTorrent employee who declined to be named in this story because of fear of repercussions from his current employer. Not only did people not want to buy movies from BitTorrent, but minimum guarantee provisions in its contracts with Hollywood also forced the company to fork over millions of dollars for content no one wanted to buy. “It kind of worked out for the studios,” the former employee says.

“There were gotchas [in those] deals that made them essentially unworkable,” Cohen agrees. “The big copyright holders were perfectly happy with where things [were], and basically didn’t want to create the streaming systems that we have today.”

BitTorrent closed the download store less than two years after its launch while also laying off half of its staff amid a recapitalization. In the years that followed, the company kept trying to figure out other business models, from P2P CDNs to decentralized chat clients — only to hit roadblocks at every turn.

The big issue was always piracy. BitTorrent successfully avoided legal liability, but it couldn’t get rid of its outlaw image. That made it impossible to strike deals that were open to others, Cohen says. “When you’re doing slightly sketchy things, the world doesn’t want to give you a nice off-ramp.”

“I never should have brought on anybody else at all. I could have simply gotten enough money off of doing the slightly sketchy stuff just on my own.”

One thing that worked was advertising. In addition to including ads directly in its products, BitTorrent would also bundle third-party tools with its installers and get a small cut when someone installed a browser toolbar or a similar app alongside the BitTorrent client. But while the ad money was paying the bills, it wasn’t meeting the expectations of BitTorrent’s investors looking for a lucrative exit.

“The [app download] numbers kept growing, but a lot of the use cases were never going to be the kind that you could legitimately either IPO or sell to another company,” the former employee says.

BitTorrent Inc.’s struggles were in stark contrast to the success of Cohen’s technology. “We were moving an exabyte of data every month,” the former senior employee estimates. BitTorrent owed much of its popularity to the wars around file sharing, which got a new poster child when Swedish authorities raided The Pirate Bay in 2006.

The battle between rightsholders and The Pirate Bay is perhaps the best-known part of the BitTorrent saga. The site was up and running again days after the raid, thanks in part to BitTorrent’s distributed architecture. The Pirate Bay didn’t run a file-sharing network, and taking its servers didn’t stop people from downloading BitTorrent’s client, or finding each other via tracker servers. All the raid took offline was a public list of downloadable torrent files, and all it took to get that back online was a recent server backup.

Undeterred, rightsholders and law enforcement continued their crusade against BitTorrent pirates and their alleged enablers, including the three cofounders of The Pirate Bay. After getting charged with criminal copyright infringement in early 2008, in April 2009 they were each sentenced to a one-year prison term and ordered to pay financial damages. Much to Hollywood’s chagrin, their site nonetheless stayed online. The exact identity of The Pirate Bay’s new operators remains a mystery to this day.

Hollywood’s fight against pirates also had repercussions for the German hackers responsible for much of BitTorrent’s tracker infrastructure. Used by a number of big torrent sites, the Denis.Stalker tracker attracted around 100 million daily peers at its peak, which were trading a million files every single day, according to Engling.

“When you’re doing slightly sketchy things, the world doesn’t want to give you a nice off-ramp.”

One day, the Berlin police’s economic crimes unit raided the apartment of one of the three people involved in running the tracker server, confiscating every piece of equipment they could find, according to Engling. “They couldn’t imagine that anyone was running something like this without making any money with it,” he says. However, money was the last thing on the trio’s mind. In the same aforementioned talk, held during the Chaos Computer Club’s annual conference between Christmas and New Year’s, they explained their motivation with three simple words: “Because we can.”

Charging them for knowingly facilitating copyright infringement wasn’t possible either, as tracker servers simply match users by hash values and never see the actual files being traded. Ultimately, the case was dismissed, according to Engling.

“Those were the wild early days,” he says.

Things changed rapidly in subsequent years. Unable to put an end to torrent sites and trackers, rightsholders shifted their focus to end users. In some jurisdictions, including Germany, this can be extremely lucrative. Law offices can effectively send cease and desist letters that double as invoices for hundreds of euros per traded file to alleged offenders, threatening them with full-blown lawsuits if they don’t pay up. Germany’s Stern magazine estimated that local file sharers received claims for around €120 million in 2010 alone.

That led to end users embracing VPNs to protect themselves — and torrent sites cashing in on that trend by adding ads for privacy services to their sites. With that, torrent sites were starting to become a major moneymaker. One 2014 study estimated that torrent sites generated around $114 million in ad revenue that year. In addition to VPNs, torrent sites have long been hawking anything from porn to dating to online casinos, with the occasional malware download thrown in for good measure.

These days, some sites also use elaborate schemes to hijack otherwise legitimate advertising and affiliate revenues. One such trick: When someone uses a torrent site, their browser opens an Amazon page for an extremely random product in the background. Toothpaste, maybe, or perhaps a totally unremarkable sweater. The goal isn’t to sell those specific products, but to plant a cookie on the visitor’s machine. If they order anything from Amazon in the following 24 hours, the pirates earn some affiliate revenue — a tactic known as cookie stuffing.

“It’s whatever the latest scam is,” says Engling, sounding disillusioned about what has become of the BitTorrent world he helped create. “The idea of democratization got lost,” he says.

But while Torrent sites were making bank, BitTorrent Inc. continued to struggle. “I did a good enough job of keeping the wheels from falling off the bus, keeping our core product working,” Cohen says. Efforts to move beyond file sharing, however, kept failing.

That included Cohen’s next big bet: For years, he worked on adapting the company’s tech for low-latency livestreaming. BitTorrent Live, as the initiative was called, even shipped consumer apps, and the company was briefly working on building a kind of live online TV service. In 2017, the company pulled the plug on those efforts and laid off almost the entire BitTorrent Live team. “This is part of the world being out to get you,” Cohen says, clearly still bitter that the project wasn’t given a chance. “That was a product that worked great and was ready to go, and was just outright fucking killed.”

“They couldn’t imagine that anyone was running something like this without making any money with it.”

Soon after, BitTorrent Inc. got sold to controversial crypto entrepreneur Justin Sun, who went on to shutter the company’s SF office after a few very tumultuous years and now operates the company from overseas. (BitTorrent did not respond to multiple requests for comment.) “It eventually got an even sketchier exit than it had been all along,” Cohen quips.

Looking back, Cohen believes that many of BitTorrent’s troubles could have been avoided if he hadn’t gone down the typical path of a VC-backed startup. Perhaps he should have just done his own thing — not unlike the path Engling chose, except more focused on modest revenue and less on free beer.

“I never should have brought on anybody else at all,” Cohen says. “I could have simply gotten enough money off of doing the slightly sketchy stuff just on my own.”

How many people still use BitTorrent today is unknown. The days of it making up a third of all internet traffic are clearly over, with one estimate suggesting that it now contributes less to residential upstream traffic than iCloud and FaceTime. BitTorrent Inc., the company cofounded by Cohen and now owned by Sun, claims that its clients still have 54 million monthly users. Popular third-party clients like Transmission, BiglyBT, and qBittorrent likely attract millions more.

I know what you download, a website that tracks usage across multiple torrent sites, estimates that about 0.25 percent of all internet users download torrents on any given day. “The number of torrent users [has been] pretty stable over the last eight years,” says the site’s admin, Andrey Rogov. Data Rogov shared with The Verge shows Russia, where Netflix and other streaming services aren’t legally available due to sanctions related to the Ukraine war, as BitTorrent’s biggest market. The United States ranks second.

Engling doesn’t have any data on the total usage of Opentracker, but he does at times look at the publicly available reporting dashboards of popular torrent trackers running his software. That data suggests we may be seeing a resurgence. “In absolute numbers, it has gotten more,” he says.

“It’s whatever the latest scam is. The idea of democratization got lost.”

One reason? Streaming is getting awfully expensive these days. For a while, people were fine paying for Netflix and Disney Plus instead of downloading movies and shows for free. “At some point, it flipped,” Engling says. “An increasing number of consumers who previously paid are dusting off their pirate hats.”

Amid that resurgence, he still sometimes tinkers with Opentracker’s code, reenergized by new technical challenges. And just last month, a Spanish anime fan took him out for beers in Berlin to thank him for his contribution to BitTorrent’s global infrastructure.

Cohen, on the other hand, is more than happy to leave that chapter of his life in the past, and instead focus on his blockchain startup Chia Network. “I don’t really care to revisit BitTorrent too much,” he says. Except, if you start a revolution, some of it always stays with you, whether you like it or not. What haunts him to this day are the struggles that came with turning BitTorrent Inc. into a venture-funded startup, and ultimately a profitable business.

“You know how people say they wake up their whole lives with nightmares that it’s the morning of a final exam in grad school? I never had those,” Cohen says. “I do get nightmares about board meetings.”

Libby will filter out AI content, kind of

2026-06-30T08:41:41-04:00

This is Lowpass by Janko Roettgers, a newsletter on the ever-evolving intersection of tech and entertainment, syndicated just for The Verge subscribers once a week.

“AI is the new frontier for us,” says Marc DeBevoise, who took over as the new CEO of OverDrive last week. OverDrive is best known for the ebook lending app Libby that is available through tens of thousands of public libraries. Like the rest of the digital publishing industry, it’s poised to face massive disruption from a huge wave of AI-generated books.

To prepare for the AI onslaught, Libby is now getting ready to introduce AI content controls, allowing readers to select in the app’s settings whether they want to see AI-generated content or not. This includes not only AI authorship, but also AI-narrated audiobooks, machine translation, and AI-generated art. “We need to tell people what’s available [and] how it was created,” DeBevoise says.

With the app’s new AI filters, OverDrive tries to strike a middle ground between allowing readers and librarians to opt out of AI and embracing what DeBevoise thinks are the technology’s upsides in areas like content recommendations and localization. (Libby first introduced some AI features of its own last year to help with book discovery, and subsequently faced some backlash.)

“AI is going to add some benefits,” DeBevoise argues. “If you think about it from an access to information and content perspective, it really does feel like a positive development, as long as it’s used in the right way.”

Much of OverDrive’s history, and its catalog, precedes today’s AI challenges. The company was founded 40 years ago to digitize books for distribution on floppy disks and CD-ROMs. It first began offering ebook lending in partnership with local libraries in the early 2000s, launched Libby as a consumer-facing app in 2017, and is now working with 92,000 public libraries, schools, and universities in over 115 countries.

Participating public libraries allow their patrons to borrow ebooks through OverDrive’s Libby app for free. Libby’s catalog consists of over 6 million books, which have been borrowed over a billion times. Most of these books were published before the emergence of modern-day LLMs. “Everything before 2020 [or] 2022 is by definition not AI,” DeBevoise says. “We know for a fact that the majority of our catalog is not.”

However, that could change quickly. Amazon began restricting the number of books self-published authors can upload per day in 2023 to combat AI slop. Kobo CEO Michael Tamblyn revealed last month that Kobo rejects nearly half of all self-published books over AI concerns. “We are in front of a firehose,” Tamblyn said.

OverDrive doesn’t offer authors a way to directly upload their own books the way Kobo and Amazon do. However, it does work with self-publishing intermediary Draft to Digital that also supplies self-published books to most digital storefronts, including Apple Books and Google Play Books. The service allows AI-generated books as long as they have gone through “extensive editing from a human,” making it all but inevitable that some AI titles find their way into Libby’s catalog. Still, OverDrive decided against using an AI checker to label books as AI-generated and is instead relying on publishers self-labeling their works via standardized metadata.

DeBevoise argues that AI can ultimately help lower barriers to information access. One example: localizing audiobooks. “It’s a great opportunity for taking domestic into international and international into domestic,” he says.

OverDrive has seen huge growth for audiobooks in recent years. Despite only being 15 percent of Libby’s catalog, audiobooks are now responsible for roughly half of all usage of the app. “It is the modality of choice,” DeBevoise says.

DeBevoise still prefers audiobooks read by actual voice actors over completely synthetic machine narration. “Nothing replaces the actual human touch on those things, and it’s not that outrageously expensive to be recording audiobooks,” he says. “But I do think putting it into a dozen or 100 languages becomes cost-prohibitive.”

AI critics have noted that AI translation, especially for literary works, can be problematic as well. Libby’s new AI filters do include an option to filter out machine translation, but that only works if books are being labeled correctly.

The Verge is concluding its weekly syndication of Lowpass this month. To keep reading Janko’s newsletter every week, subscribe here.

How the World Cup became a US streaming success story

2026-06-25T11:24:55-04:00

Guillermo Ochoa #13 of Mexico is tossed into the air by his teammates after the team's 3-0 victory in the FIFA World Cup 2026 Group A match between Czechia and Mexico at Mexico City Stadium on June 24, 2026 in Mexico City, Mexico. | Getty Images

This is Lowpass by Janko Roettgers, a newsletter on the ever-evolving intersection of tech and entertainment, syndicated just for The Verge subscribers once a week.

The 2026 World Cup is breaking streaming records around the world: Brazil’s CazéTV YouTube livestream of that country’s opening game against Morocco surpassed 12 million concurrent viewers, a new milestone for YouTube. The South Korea versus Czech Republic game was streamed by 3.86 million viewers in South Korea at its peak, doubling a previous record set by BTS. And the BBC clocked more than 600,000 concurrent streams during the France-Senegal match-up, making it the biggest-ever live event streamed in 4K for the broadcaster.

Even in the United States, where soccer traditionally isn’t as widely watched as other types of sports, the Cup seems to be striking a chord. The Mexico vs. South Korea game attracted an audience of 6.1 million viewers to Telemundo’s streaming platforms, making it the largest Spanish-language soccer stream shown to US audiences to date. And the opening game of the United States team attracted 1.1 million streaming viewers on Tubi alone, turning it into the most streamed English-language match of the team in World Cup history.

The stateside success of the World Cup is the direct result of a series of unlikely business decisions, legal challenges, and regulatory pressures — a kind of perfect storm for the beautiful game.

Hot summer nights

A key decision impacting the 2026 World Cup was made in 2010, long before sports streaming was really on anyone’s radar.

Back then, FIFA awarded the 2022 World Cup to Qatar, a country where summer temperatures often exceed 104 degrees Fahrenheit. When the soccer association considered moving the tournament to the winter, US broadcaster Fox threatened to sue because the change in schedule impacted its existing sports programming.

This led to FIFA striking a sweetheart deal with Fox and Telemundo for the 2026 rights in 2015, reportedly without considering any competing bids. Had Disney and Univision been part of the bidding process, the rights likely would have cost a lot more, and Telemundo owner NBCUniversal could have walked away empty-handed.

But thanks to that no-bid deal, brought about by Qatar’s scorching summer temperatures, NBCUniversal is now offering a Spanish-language feed of every game on its Peacock streaming service for just $11 a month.

Soccer is catching up

Speaking of Spanish: The fact that US soccer fans have two competing options, Fox and Telemundo, to watch games live harkens back to the long-held belief that Americans don’t care about soccer. Or, more specifically, non-Hispanic Americans don’t care about soccer. (Nearly three-quarters of U.S. Latinos call themselves soccer fans; 22% of Hispanics identify as “superfans” as opposed to only 7% of non-Hispanics.)

It’s true that World Cup viewing in the US has historically lagged behind other major sporting events. This year’s Super Bowl was watched by more than 125 million viewers, while fewer than 26 million Americans tuned into the 2022 World Cup finals.

However, the population of the country is changing, and rapidly so. There are now 68 million Latinos in the United States, up from 35 million in 2000. Add the massive success of the US Women’s soccer team on the world’s stage and everything it has done for the sport, and things are starting to shift: According to a 2025 survey, soccer is now more popular than baseball in the United States.

All this could lead to a scenario in which US broadcast rights for the World Cup become a lot more expensive. Some even speculate that FIFA may start selling English- and Spanish-language broadcast rights in a single package going forward, which could make the tournament a lot more expensive for US broadcasters and viewers alike.

The Venu debacle

Viewers who prefer to watch the Cup in English also have a relatively affordable option available to them: Not only does Fox’s Tubi service stream select games for free, the company’s Fox One streaming service offers access to every single game for $20 a month — which wasn’t at all how Fox envisioned to price this tournament.

In February 2024, Fox, Warner Bros. Discovery and Disney announced plans to launch a joint sports streaming service. Known as Venu, the streamer was supposed to combine sports programming from ESPN, Fox Sports, and networks like TNT and TBS under one roof.

Venu was good news for hardcore fans frustrated by the growing fragmentation of sports streaming, with games even from the same leagues often spread out across multiple paid services. However, combining all that programming came with a hefty price tag: Venu was supposed to cost $42.99 per month at launch.

Those plans fell apart due to regulatory pressure and a lawsuit by live TV service Fubo, with Venu’s joint owners pulling the plug on the venture in early 2025. Fox instead decided to go it alone, launching Fox One as a sports streaming service a year ago. That decision effectively cut the price soccer fans have to pay to tune into English-language livestreams by more than 50 percent.

Paid streaming beats pirates — for now

Fox streamed the entire 2022 World Cup on demand for free on Tubi. When the broadcaster announced plans to move most games behind the paywall this time around, I initially assumed this would lead to a familiar pattern: Faced with steep subscription fees, sports fans instead opt to pay a few bucks for a VPN that allows them to tune into the livestreams of foreign public broadcasters like the BBC or Germany’s ARD.

I clearly wasn’t the only one with that assumption. ExpressVPN is an official World Cup sponsor, and other VPN providers have been aggressively marketing their services to soccer fans as well.

However, early data suggests that those messages aren’t landing with consumers. US download numbers for VPN apps have actually been below their base level since the beginning of the World Cup, according to data app analytics startup Appfigures shared with me for this story.

Subscription video and live TV apps like Peacock, Fox Sports, and Tubi saw their download increase by 150 percent since June 11, according to Appfigures. And on Google Play, Fox One, and Peacock currently rank among the Top 5 most downloaded apps.

All this seems to suggest that, at least for now, paid services are winning. And while it took a series of unlikely events to get to this point, broadcasters hopefully take away the right lesson: Price your services right, and people will actually pay to watch every gooooooal.

No more lightbulbs, much more sports: Five predictions for Roku’s future

2026-06-17T16:10:26-04:00

This is Lowpass by Janko Roettgers, a newsletter on the ever-evolving intersection of tech and entertainment, syndicated just for The Verge subscribers once a week.

When Fox announced its acquisition of Roku earlier this week, executives of both companies were quick to promise that not much would change in the near future. Sure, getting its hands on Roku will help Fox become a major force in streaming, and surpass the viewership of Netflix in the United States when you include Fox’s TV networks. But Roku will remain open to all streaming services, and Fox will keep selling its programming to anyone, they pledged.

Of course, that doesn’t mean everything will stay the same. I expect that the deal will have a significant impact on both companies, and that Roku owners will see major changes once it closes. Here’s what I believe is in store for Roku’s future:

The Roku Channel will become a Roku exclusive

One of the questions analysts had on a joint Fox and Roku investor call earlier this week was: What does this deal mean for the future of Tubi and The Roku Channel, the ad-supported streaming services operated by the two companies? The diplomatic answer: The services are complementary, with Tubi mostly serving up movies and TV shows on demand while The Roku Channel has a lot more eyeballs on free, ad-supported streaming channels (also known among industry insiders as FAST channels).

That’s true, to a degree. Tubi’s leadership initially didn’t believe in the potential of linear channels and instead put a greater emphasis on being a free Netflix alternative. Roku, on the other hand, had its software deeply embedded in smart TVs and knew that it could build a replacement for the cable guide if it went all in on linear streaming channels.

At the same time, Tubi and The Roku Channel do squarely compete with each other on third-party platforms like Fire TV, Google TV, and Samsung, where consumers can download either app to access free programming. Roku expanded onto those platforms a few years ago, but that expansion has been an ongoing cause of confusion for consumers. Many simply can’t understand why the Roku app they downloaded onto their Samsung TV or Fire TV doesn’t offer them the exact same experience as a Roku device.

Another cause of frustration: Roku offers owners of its devices the ability to subscribe to third-party services like HBO Max and Paramount Plus through The Roku Channel. But if they fire up The Roku Channel’s app on the Samsung TV in their bedroom, those paid subscriptions are MIA.

To simplify all that, I predict that Fox will opt to make The Roku Channel a Roku exclusive, and bet on Tubi as the sole app for ad-supported viewing on third-party devices.

Roku will abandon its IoT efforts

Over the years, Roku has tried many things to launch an ecosystem of devices around its platform. There was that streaming-optimized mesh router that never launched, and the smart speakers that have been sidelined in recent years.

More recently, Roku has bet aggressively on low-cost cameras, lightbulbs, and alarm systems. All of that made sense when the company was trying to go it alone and compete with the likes of Google and Amazon. The fear was that consumers may choose competing products if they couldn’t access their security camera feed on their Roku TV.

For Fox, IoT is little more than a distraction. The margins of these products are just too low, and there’s easier ways to get people to watch more Fox programming than launching yet another smart plug. That’s why I predict that Roku will phase out its IoT products and instead embrace the Matter standard to more easily work with third-party hardware.

Howdy will become an unexpected hit

Fox has been a bit of an outlier in the streaming wars. Companies like Disney and Warner Bros. Discovery spent tens of billions of dollars to build paid streaming services to catch up with Netflix. The Murdoch empire instead waited on the sidelines and bet on Tubi as a free streaming alternative.

Fox Nation aside, the company’s first real paid streaming service has been Fox One, the sports streamer that launched less than a year ago. That wait-and-see approach is very similar to Roku, which long emphasized ad-supported streaming over paid subscriptions. The first paid service Roku built in-house is Howdy, a low-cost streamer catering to budget-conscious consumers.

At $2.99 per month, Howdy is unlikely to become a major moneymaker for Fox. However, I predict that Fox may treat Howdy as a bit of a wild card bet: one more way to monetize its content and a good opportunity to needle and undercut the competition. With that strategy, and with access to Fox’s massive back catalog of popular TV shows, I predict Howdy may turn into an unexpected hit with consumers wary of too many ads.

Roku will double down on international expansion

Fox gave up on some of its international ambitions when it sold its stake in Sky to Comcast in 2018 and shed more assets abroad when Disney bought 21st Century Fox the following year.

Since then, Fox has made some efforts to expand abroad again with Tubi, which is now available in Canada, Australia, and Latin America. Those happen to be the very markets that Roku has also been targeting for its own international expansion, with some success: Roku claims to be the number one streaming platform in Canada and Mexico, and its hardware partners are now selling Roku-branded TVs in over 15 countries.

However, at its core, Roku, like Fox, is still a US-centric business. In its 2025 annual report, the company disclosed that over 90 percent of its devices revenue still comes from the United States. The same is true for advertising and services revenue, with the report stating that “we currently generate the vast majority of our revenue in the United States.”

Roku has been cautious to ramp up international expansion, especially in Europe, where Amazon has a much stronger foothold with Fire TV, to keep costs in check. With Fox’s support, I expect more aggressive moves into foreign markets along with Tubi. Both Roku and Tubi can help build scale for Fox’s streaming efforts. Fox can cash in on those efforts once it has millions of consumers hooked on free programming and affordable TVs.

Sports will be front and center on Roku

It’s hard to ignore the Fox One sports streaming service these days, especially if you’re trying to keep up with the World Cup. I expect Fox to double down on that momentum and tightly integrate Fox One with Roku. This doesn’t mean that Fox One will become a Roku exclusive. Instead, you’ll find sports programming, and deep links to the sports streaming service, everywhere, including Roku’s newly redesigned homescreen and its Sports Zones.

In other words: Don’t be surprised to see Fox build a stadium in Roku City.

Is Apple TV the new HBO?

2026-06-11T08:19:25-04:00

This is Lowpass by Janko Roettgers, a newsletter on the ever-evolving intersection of tech and entertainment, syndicated just for The Verge subscribers once a week.

Severance. Pachinko. Silo. Ted Lasso. Over the past couple of years, a number of Apple TV shows have become hits with audiences and critics alike. And yet, compared to the size of other subscription services, Apple TV still barely makes a dent.

In Nielsen’s most recent The Gauge report, Apple’s service failed once again to make the top 10 list of most-used streaming services, suggesting that its audience is smaller than not just that of Netflix and Disney Plus, but also Tubi, HBO Max, and The Roku Channel.

New data released by subscription insights startup Antenna last week suggests that this may not be a contradiction at all: Engagement with Apple’s video services is heavily driven by a few anchor shows, according to Antenna’s new State of Subscriptions report, whereas Netflix viewing is spread out much more across a wider range of titles.

The report estimates that 32 percent of what the company classifies as heavy viewers watched the Apple TV show Shrinking in March of this year, while 31 percent watched the action series Monarch: Legacy of Monsters. On Netflix, only War Machine was watched by 25 percent of heavy viewers, while all other titles remained below the 20-percent mark.

“Big scripted hits anchor heavy viewership on Apple TV,” the report notes. “Netflix Heavy Viewers spread across the slate, signaling breadth of engagement rather than concentration.”

A few tentpole titles that get people hooked and become cultural moments: That’s a model premium cable networks like FX, HBO, and Showtime relied on for decades. It’s also something a number of video subscription services briefly tried to replicate in the streaming age as they aimed to compete with Netflix.

When HBO Max first launched in 2020, it was all about extending the HBO brand and curation approach into streaming. Then, WarnerMedia merged with Discovery, which brought a bunch of reality TV shows to the service, muddying the value proposition. The service was even briefly rebanded as Max to attract a broader audience, but Warner Bros. Discovery reneged on that change last year.

Likewise, Disney tried to give its streaming operations more of a premium TV network feel when it launched the FX on Hulu brand with great fanfare in early 2020. Cashing in on the success of the cable network, FX on Hulu was supposed to not just host shows that had previously aired on cable, but also FX-branded Hulu originals. However, it turned out that establishing a premium brand within a service that had a little bit of everything was easier said than done. The company has since phased out the “FX on Hulu” branding and simply uses FX as a content category alongside others, like ABC, Hotstar, and Freeform.

Which raises the question: If being the HBO of streaming didn’t work for HBO, can it work for Apple TV?

“Super-serving against a niche audience you understand really well is a business model,” says Paul Pastor, the chief business officer of Quickplay, which builds streaming apps and infrastructure for media companies like MSG and Gray Media. “They’re doing that extremely well with their shows,” he adds.

Some of the engagement differences between Apple TV and Netflix can be explained by catalog size. Apple TV currently offers subscribers access to around 220 shows, according to JustWatch, while Netflix’s catalog contains around 3,300 shows.

However, Pastor also believes that Netflix is a lot more proactive about surfacing lesser-known gems to its audience. “They really figured out the formula of how to manage that through their recommendations and personalization engines,” he argues. “They’re fundamentally focused on their tech platform.” (Pastor’s company is pitching AI-powered personalization solutions to streaming service operators.)

Pastor acknowledges that Apple does have a few key advantages that allow the company to stick to an HBO-like model for its streaming service.

For one thing, Apple does have a different business model than much of its competition. Apple TV is the only major streaming service that doesn’t have an ad-supported tier yet, and Apple as a company has a huge devices business that allows it to pursue different priorities. This includes using Apple TV as an entry point to then resell other services.

Case in point: If you turn on your Apple TV or browse the Apple TV app on your iPhone, you may find those tentpole Apple TV shows, but also third-party content from services like Hulu, HBO Max, or Prime Video. And when you subscribe to those services through Apple devices, the company grows its already massive services business.

“At the end of the day, Apple is also an aggregator,” Pastor says. “That might be the best business model for them.”

The best business model for now, that is. If the company were to embrace advertising for Apple TV, like it has been rumored for some time, it may embrace a more Netflix-like model with a broader catalog, more personalization, and less of a focus on tentpole titles. Not only do advertisers demand scale, but an ad-supported tier would also fundamentally change the incentive structure for Apple. Right now, the company makes money with monthly fees, no matter how much each subscriber watches. With an ad-supported tier, every additional ad break shown generates extra revenue.

Until that happens, Apple TV may just be the next best thing we have to HBO. Sorry, HBO Max.

The secret to Roku’s success: not being cool

2026-05-28T09:56:41-04:00

This is Lowpass by Janko Roettgers, a newsletter on the ever-evolving intersection of tech and entertainment, syndicated just for The Verge subscribers once a week.

Roughly 10 years ago, someone told me that Roku was making “cheap hardware to sell to Walmart customers in flyover states.” The remark was meant to be an insult, belittling a company that seemed to care more about hardware profit margins than design and innovation.

Still, I’ve been thinking about it a lot over the years. And as Roku became a major force in streaming hardware, surpassing 100 million households last month, I’ve come to the conclusion that Roku’s secret superpower may just be that it embraced not being cool.

Case in point: the new homescreen the company began rolling out this week. It’s a refresh that had been overdue for some time, and, in many ways, it’s the bare minimum the company could have done — which means it’s likely going to be a huge success.

Content-forward, but not overbearing

Companies that make streaming dongles and smart TVs have long pushed the idea of content-forward user interfaces. The gist of it: People don’t want to launch an app, browse through row after row of thumbnails, and then do the same thing all over again in the next app if they can’t find something they want to watch right away.

Content-forward interfaces instead serve up personalized recommendations for individual movies and shows directly on the homescreen. Add some tiles to quickly access the next episodes of your current favorites, and you’ve got an interface that makes it much easier to actually watch TV without wasting time on discovery.

Amazon was first to embrace a content-forward interface with its Fire TV devices more than a decade ago. Google followed suit with Android TV, and then doubled down on the idea when it relaunched its living room platform as Google TV. Today, virtually every smart TV platform has some kind of content-forward interface.

The sole exception, up until now: Roku. The company added some dedicated content categories to its sidebar over the years, but largely kept the homescreen a bare list of app icons.

This week, that’s changing, but in a very Roku-like way. “I don’t think ours looks like those others,” says Preston Smalley, who led the redesign as Roku’s VP of viewer product. “Yes, it has content on it. Yes, it has a destination that you can browse, but I think it still stands apart as something unique.”

While others lead with big, wall-to-wall hero art to highlight individual shows (and the occasional ad), Roku’s new homescreen starts with a comparably small, personalized “top picks for you” section, followed by rows of tiles that look a lot like the old Roku homescreen: apps, destinations, shows, all arranged in a tight grid, accompanied by a retractable sidebar, and with enough room for a big ad on the other side (like its competition, Roku is also looking to make more money with its homescreen).

It all looks very predictable and familiar. Boring, perhaps, but also less distracting than the bigger and bolder interfaces used by Fire TV and Google TV with their massive banners and autoplay trailers. Roku’s existing customers will feel right at home.

The logical conclusion of cord-cutting

Roku has been on the forefront of one key trend: While some of its competitors were trying to make nice with cable companies, Roku embraced cord-cutting from day one. The company also realized early on that cord-cutting was primarily about saving money. It priced its hardware accordingly, struck partnerships with budget consumer electronics makers like TCL to expand into smart TVs, and has been heavily betting on free, ad-supported TV for years.

Roku executives also early on predicted the logical conclusion of this trend: If all TV is going to be streamed one day, then everyone will need to be able to access streaming — including tens of millions of people who aren’t exactly tech-savvy.

Smalley knows this late-adopter world well. Before Roku, he worked nearly a decade at Comcast. Doing user research for the cable company was an eye-opener. “I’d go to people’s homes,” he recalls. “Some people would have remote controls where they taped over the buttons and tried to simplify it down to what looks more like a Roku remote.”

Those attempts to dumb down devices continue in the streaming age. “You’ve got people that maybe only use a few apps. Maybe their son comes by and configures it so that all their apps are right there. Unfortunately, not everyone’s got a loving son. For each of those, there’s 10 more that don’t have that.”

That’s why the new Roku homescreen has a dedicated “quick access” section of frequently-used apps above the fold. The section auto-updates based on usage, and can also include the tile to change your HDMI port if you’re one of those people who frequently switches to a game console. “You have different people that are looking for different things,” Smalley says. “And yet you want to have one experience that works [for everyone].”

Roku City may be coming to mobile

One noteworthy feature of Roku’s new homescreen is a section called Your Daily Scoop that aims to serve up zeitgeist-influenced viewing recommendations. Think awards, holidays, pop culture moments, and yes, even the news — with some guardrails. “We’re not going to talk about the war,” Smalley says. “It’s definitely a curated experience.”

It’s still essentially running on autopilot, so there may be slip-ups. More importantly, it does highlight one of Roku’s weaknesses: a lack of first-party data that doesn’t originate from people using Roku’s devices. Google uses search trends to inform what it shows on Google TV, and Amazon can presumably tap into Alexa usage patterns to personalize Fire TV.

Roku doesn’t have any of that, and is trying to make up for it with third-party data. “We’re using some of the LLMs [to figure out] what people are looking for,” Smalley says.

Still, Roku also does have a few things the competition lacks. One example: Roku City, the company’s screen saver that became a cult hit during the pandemic, now has its own app icon. “We wanted to provide a way for [people] to launch it [without having to] wait 10 minutes before this thing comes on,” Smalley says.

It’s a small step, but one that’s in line with the company’s efforts to transform Roku City into both a destination and a moneymaker. In recent years, it has integrated ad campaigns, minigames, and even a live concert into the purple-tinged cityscape. Next up could be an expansion to mobile. “We have 34 million people using our mobile app every month,” Smalley says. “We’re looking at the right way to bring some of Roku City to mobile.”

If and when that happens, it will likely be deceptively simple — and incredibly successful.

AI video is moving beyond clip slop

2026-05-21T12:19:36-04:00

Luma/X" data-portal-copyright="Image: Luma/X" data-has-syndication-rights="1" src="https://platform.theverge.com/wp-content/uploads/sites/2/2026/05/Screenshot-2026-05-21-at-10.55.11AM.png?quality=90&strip=all&crop=0,0,100,100" />

A still from Innovative Dreams, a new production company by Luma and Wonder Project | Image: Luma/X

This is Lowpass by Janko Roettgers, a newsletter on the ever-evolving intersection of tech and entertainment, syndicated just for The Verge subscribers once a week.

Hollywood is cooked — or so a growing number of people on social media would like you to believe. Their purported proof: AI-generated clips of Daniel Craig riding a Vespa through an Italian city, Godzilla fighting King Kong, or The Avengers zooming through Manhattan.

In reality, cheap slop like this won’t replace Hollywood blockbusters any time soon. However, a new generation of AI video solutions could upend how studios work. That’s because, until recently, AI companies basically tried to sell Hollywood on the same idea as those Twitter guys, with a slightly more palpable spin. The pitch, in a nutshell: AI video will allow everyone to make movies faster, cheaper, better — one prompt at a time.

“The premise was: Substitute your camera for our video model,” says Luma AI CEO Amit Jain, whose company used to make that very same pitch to studios. But when it began partnering with the entertainment industry, it received a crash course in the way Hollywood actually works.

“It’s not sufficient to just produce a clip,” Jain says now. “Because then what?” Clips generated by video models are typically 10 to 16 seconds. “That’s not a shot. That’s not a sequence. That’s not a scene,” Jain says. “Churning out short videos is not enough.”

Now, AI companies like Luma believe they have found a better way to sell Hollywood on AI. The gist? Don’t just use AI for video — use it for everything.

Luma has been working on AI agents that can help with the entire production process. Jain compares this transition to the way software development with AI has evolved, with companies like Anthropic moving from simple vibe coding to agentic workflows.

“It’s not sufficient to just generate a little bit of code,” Jain says. “We need these systems to do long-horizon, end-to-end work. That’s what solves problems for people.” AI agents can do the same for Hollywood, he believes.

Luma isn’t alone with that approach. Just this week, Google unveiled a new version of its AI media authoring platform Flow that also emphasizes agentic end-to-end work over simple clip generation. “There’s this huge evolution that’s happening in generative tools,” says Google Labs VP Elias Roman. “Moving forward, they’re going to become much more like agents.”

In the new version of Flow, an agent guides the user through multiple steps, from starting with a concept to fleshing out plotlines to developing characters to setting the desired look and feel. And when it’s ultimately time to generate video, the agent uses the things it learned along the way to achieve a specific outcome without having to be prompted about every single detail.

One issue this is aiming to solve is consistency. Generative AI has long struggled with keeping characters looking the same from clip to clip. In the new version of Flow, users can add a character they developed for a project to a prompt simply by tagging it, just like you’d add a colleague to a Slack conversation.

A new generation of video models is also better at understanding physics, the look of a certain era, and cinematic languages. Google’s Flow is powered by the company’s new Gemini Omni world model, while Luma has developed Uni-1 as a unified model that doesn’t need extremely complex prompts anymore to make sense of an envisioned world.

Luma recently teamed up with Amazon to produce The Old Stories: Moses, a companion special for MGM’s House of David show. While shooting Moses, actors would perform in front of LED walls showing backgrounds generated with Luma’s video models, while their costumes were rendered with AI as well.

Announcing Innovative Dreams, a new production company by Luma and Wonder Project built by filmmakers for filmmakers, and our first production — Moses starring Sir Ben Kingsley, coming this Spring on Prime Video. AI is a fundamental change to the craft and business of filmmaking,… pic.twitter.com/Rx4K3aou9B
— Luma (@LumaLabsAI) April 16, 2026

If a shot turned out to look not quite right, all it took was one new prompt to generate a new asset. “This level of production would take about six weeks to eight weeks per hour of television,” Jain says. “Now, it’s taking them a week.”

Some studios are increasingly embracing that change. Netflix acquired Ben Affleck’s AI company InterPositive in March, and launched its own AI animation studio the same month. Two major Hollywood studios already use Luma’s AI agents, Jain claims. He declined to name names, but the company has been publicly celebrating some smaller wins: Luma recently announced the launch of a joint venture with indie studio Wonder Project, which made Moses.

These developments undoubtedly will lead to job losses, even though the scale of the impact is still unknown. If studios can make a TV show in a month instead of 10, it won’t be sending out checks for those other nine months. The counterpoint AI boosters like to raise is that this will lead to more productions. This could be a silver lining for Los Angeles in particular, which has seen production days plummet in recent years.

Now, we’ll just have to see if Hollywood uses this tech for something people actually want to watch.

Netflix is building an AI animation studio

2026-05-21T10:39:55-04:00

Netflix’s job listings indicate an generative AI-heavy animation studio is coming. | Image: The Verge

This is Lowpass by Janko Roettgers, a newsletter on the ever-evolving intersection of tech and entertainment, syndicated just for The Verge subscribers once a week.

Netflix has been building a new internal studio called INKubator that aims to use AI to produce short-form animated content: The streamer is hiring for a wide variety of roles, including producers, software engineers, and CG artists to staff INKubator, according to a number of recently published job listings.

Netflix has yet to publicly announce its plans for INKubator, which job listings also sometimes refer to as INK. The company did not immediately respond to a request for comment.

A handful of LinkedIn profiles suggest the unit quietly launched in March. Its leadership includes Serrena Iyer, who previously held strategy and operational roles at DreamWorks Animation, MRC Studios, and A24 Films.

INKubator is just Netflix’s latest push to use AI for production. Earlier this year, it acquired InterPositive, an AI startup founded by Ben Affleck. But while InterPositive is primarily focused on the use of AI in post-production, INKubator appears to go much further: A listing for INKubator’s head of technology calls it “our next-generation, creative-led, GenAI-native animation studio,” with plans to “bridge innovation with imaginative storytelling.”

INKubator’s long-term technology strategy will focus on “GenAI-enabled workflows, artist tooling, and scalable, secure multi-show environments,” according to the listing, suggesting that this is about much more than one-off experiments. “We aim to develop feature-quality content,” emphasizes another listing.

At least for now, Netflix doesn’t plan to produce the next KPop Demon Hunters with AI. Instead, INKubator will be all about “creating animated shorts and specials using experimental GenAI-native production pipelines,” as one of the listings puts it.

However, at least one job listing suggests the company is already considering taking the technology beyond shorts. INKubator’s head of technology will “ensure that INK’s technology investments accelerate creative ambition […] as we ramp up activity and aim to expand into longer-form content,” a listing for that position states (emphasis added).

Netflix could potentially use AI-generated short-form content in various ways. The streamer recently revamped its mobile app, adding a TikTok-inspired vertical video feed called Clips. At the moment, this feed only includes trailers, behind-the-scenes footage, and other promotional content for its long-form programming. However, one could imagine that the feed could one day also include original short-form stories, including AI-generated shorts.

The streamer has also been making a push to establish itself as a kid-safe alternative to YouTube by bringing creators like Ms. Rachel onto its platform. Generative AI could be one way for Netflix to further scale its kids programming and compete with a flood of videos targeting kids on YouTube.

YouTube-native studios have been among the first to use generative AI for animation. Animaj, the studio that produces the popular kids show Pocoyo, has been vocal about incorporating AI into its production pipeline since 2024. Toonstar, maker of the YouTube series StEvEn & Parker, also uses AI.

However, there has also been a significant backlash against the use of AI in animation. Japanese animation legend Hayao Miyazaki famously called AI “an insult to life itself,” and labor unions representing animators from multiple countries organized a protest against generative AI at the 2025 Annecy Animation Film Festival.

Efforts to popularize the use of AI for animation beyond Hollywood have also faced setbacks. AI animation company Invisible Universe, which I wrote about last year, is shutting down its creator platform Invisible Studio by June 1st. Invisible Universe CEO Tricia Biggio told me in an email this week that her company was focusing on enterprise clients going forward.

Netflix has its own, impressive AI-powered voice search

2026-05-07T12:05:31-04:00

This is Lowpass by Janko Roettgers, a newsletter on the ever-evolving intersection of tech and entertainment, syndicated just for The Verge subscribers once a week.

A small subset of Netflix viewers just got a new way to find movies and TV shows. The streamer recently started to test an AI-powered native voice search feature with some subscribers, prompting them to press the Netflix button on their remote control to ask for viewing recommendations.

Once viewers press the button, they’re presented with a few search suggestions, including phrases like “I need a good cry,” “watch in the background,” and “help me stay awake.” Each of these suggestions leads to a set of viewing recommendations, but there’s also an “Ask” button with a waveform icon. Select it, and you’ll start an AI-powered voice search that delivers viewing recommendations in response to natural language prompts.

I’ve had access to this new voice search feature for a few days now, and found it remarkable for two reasons: While still in beta, it’s impressive in its ability to serve up appropriate and interesting viewing recommendations to even the most esoteric requests. It also squarely circumvents the voice assistants and search features built into smart TVs and streaming devices, highlighting the power struggles between TV OS platforms and the streaming services running on them.

Netflix is currently testing its new voice search feature with select members in the U.S. I got access to the feature on a Chromecast with Google TV streaming dongle and a TCL Google TV, but wasn’t able to access it on Roku or Fire TV devices. Even users who have multiple devices running the same smart TV OS have reported that they were only able to access it on a subset of those devices.

Playing with the feature for a few days, I was impressed by its ability to find relevant picks for a wide variety of requests, ranging from the rather obvious (“date night movies from the ’80s” served up The Breakfast Club) to the very specific (“I like the music of Brian Eno, what should I watch” resulted in a suggestion for Abstract: The Art of Design).

I stress-tested the feature more than people usually would during day-to-day use, but it was able to make solid recommendations in response to unusual requests. I honestly thought asking for “fun kids TV shows about death” wouldn’t get me any results. Instead, it served up A Series of Unfortunate Events and Raising Dion, which both were on point.

“I had too much coffee today, what should I watch” resulted in Netflix suggesting a “laid-back” Sheng Wang comedy special and the Headspace Guide to Sleep. When I simply said “hurt people hurt people,” the Netflix app responded with “Hurt people do hurt people. Here are some of the stories behind that,” followed by recommendations for 13 Reasons Why and It’s Okay to Not Be Okay.

Interestingly, Netflix’s voice search doesn’t appear to be tapping into the service’s personalization engine right now. When you ask for recommendations based on what you’ve recently watched, the app’s response is “We can’t answer that one yet, but we’re working on it!” (Speaking of which: All responses are shown as written text. Netflix’s voice search doesn’t use any voice output of its own, which I didn’t mind at all.)

Responses are often accompanied by suggestions to further narrow your results, including “more unhinged” and “more bittersweet.” However, narrowing down recommendations this way doesn’t always work. A search for Cantonese comedies yielded many relevant results. But once I followed Netflix’s suggestion to narrow those results by pressing the “more quirky” button, the streamer suddenly tried to sell me BoJack Horseman as a “Cantonese comedy with a quirky twist.”

There are some things that can trip up Netflix’s voice search. When I asked for Blaxploitation movies, the app’s voice recognition system instead heard me ask for “Black exploitation movies,” and declined to provide me with any results. Which, yeah, probably the right choice.

A request for “TV shows about porn” was also declined, but Netflix did happily serve up plenty of suggestions when I requested “steamy TV shows.” That’s notably different from Google’s Gemini assistant on Google TV, which informed me that it could not fulfill that request, adding passive-aggressively: “I can help you find appropriate shows if you’d like.” That alone shows why a service like Netflix might want to have its own voice search instead of relying on the one provided by your TV or streaming device maker.

Another major reason for Netflix to build its own voice search: The streamer wants viewers to remain in its own app, and exclusively recommends Netflix content. Smart TV OS platform operators like Google, Roku, and Amazon on the other hand would like consumers to use their own, universal search that also recommends results from their own services and services from partners they can monetize.

This power struggle between platforms and publishers has been going on for years, and has also long extended into voice search. For consumers, the results have been confusing: When you press your remote’s mic button while using Hulu or Disney Plus, you’ll most likely access the universal search function built into your TV or streaming device, with results from the app you’re in intermixed with those of other publishers. Do the same while browsing Netflix or YouTube, and you’re only searching the catalogs of those respective apps.

Thanks to their market share, YouTube and Netflix have so much power that they have been able to demand voice routing privileges not available to smaller publishers. Both companies initially just used that privilege for speech-to-text functionality, essentially feeding your voice queries into the same search fields you can also access with your remote to search for a show by name.

With the emergence of LLMs, both services now extend those privileges for more full-featured voice assistance. YouTube recently launched its own conversational AI tools, allowing viewers to access Gemini-powered voice features on devices made by Amazon, Roku, and others.

Just like YouTube’s implementation, Netflix’s take on AI voice search shows that consumers can benefit if apps have access to the mic. It just shouldn’t take massive market share — or a branded button on your remote — for publishers to get to build such experiences.

After working on the Vision Pro, this AR veteran is going back to phones

2026-04-30T12:58:39-04:00

This is Lowpass by Janko Roettgers, a newsletter on the ever-evolving intersection of tech and entertainment, syndicated just for The Verge subscribers once a week.

When Mark Drummond was working on the Vision Pro at Apple, he had a bit of an epiphany that didn’t really fit Cupertino’s preferred narrative.

Drummond was managing the Character Intelligence Team, which among other things built the Encounter Dinosaurs demo. Preinstalled on the Vision Pro, the demo puts viewers eye to eye with interactive prehistoric creatures.

“We built that with Jon Favreau,” Drummond says, referring to Apple’s long-standing partnership with the Mandalorian director. “Before the headset was available to us to take down to Burbank, we used iPhones and iPads,” he recalls. Relying on mobile devices for demos made sense. VisionOS is essentially an iPadOS fork. “It did actually work out. We had a really good experience with iPhone and iPad,” he says.

Through that experience, Drummond realized that these mobile devices weren’t just decent stand-ins for the headset. “What we learned looking for sources of surprise and delight with interactive characters in mixed reality is that the headset is actually not the best [device] for this kind of thing,” he says.

“I still think it’s a pretty fabulous piece of hardware,” Drummond says. However, headsets can also be alienating, and separate viewers from the world and from the people around them. “It’s kind of lonely,” Drummond says. Having an AR app on a phone, on the other hand, makes it much easier to show it to others. “People can lean in over your shoulder,” Drummond says.

That’s why, shortly after leaving Apple in 2023, Drummond embraced mobile devices for augmented reality storytelling. For the past two and a half years, he has been working on a new iPhone AR app called Pixi that’s all about mobile-first interactive storytelling.

In short, Pixi is working on something that could best be described as the AR version of the email greeting card. When it launches in the coming weeks, Pixi will let anyone pick an interactive character and a scenario, add a personalized message, and then send it to their contacts via iMessage or WhatsApp. Once the recipient opens such a Pixi message, the character appears overlaid in the camera view of their real-world environment, and interacts with them.

During a recent demo, Pixi’s AR experiences included an animated cat and a robot. They could tell jokes, play tic-tac-toe with you, or challenge you to a game of Whac-A-Mole, right on your desk.

At first glance, none of it seemed especially groundbreaking if you’ve ever played with an AR Snapchat filter or any other mobile AR app. “If you’re not slightly embarrassed by the first product, you launched way too [late],” Drummond says, suggesting that the company will incorporate additional characters and more advanced storylines in the future.

However, Pixi is incorporating some interesting interactive elements: The cat comedian, for instance, delivers one dad joke after another until the app detects the viewer cracking a smile via the phone’s front-facing camera. After that, the character thanks the viewer for their attention like a comedian ending their show and delivers a personalized greeting message.

“In order to make a character feel like it’s present, it has to pay attention,” Drummond says. That includes not just paying attention to the viewer, but also the environment they are in. “This kind of attention is only possible through on-device AI,” he says. Pixi uses AI and machine learning to recognize facial expressions, and the app also downloads custom ML models on the fly to recognize objects and then incorporate them into a story.

When mobile AR first became popular with Snapchat filters and ARKit-based apps, developers had very little to work with when it came to interacting with a viewer’s real-world environment. Basically, apps would detect surfaces and just put any virtual object on top of those surfaces. Over time, AR apps became smarter about occlusion, allowing characters to hide behind real-world objects. Eventually, some even offered rudimentary object detection.

With recent advancements in AI, there’s a potential to supercharge object recognition, to the point where apps will be able to recognize classes of objects, understand how they work, and incorporate them into interactive experiences. Here’s how Drummond envisions Pixi’s AR greeting cards will work in the future:

“Let’s say that I have a friend [who is] a lawyer. She got promoted, and I want to say congratulations. I send her a [Pixi with a] golden retriever, which appears on the office floor. It looks at her, then starts sniffing. It walks over to a filing cabinet, and is very focused on a drawer in that filing cabinet. She thinks: Dog, that’s just old tax returns. The dog will not relent. It looks, sniffs, looks winsome. She opens the drawer. Inside, we’ve hidden a huge virtual dog biscuit [with the message]: ‘Congratulations, see you Friday.’”

Would all this be technically possible with a Vision Pro app as well? Perhaps. But there are plenty of other reasons holding the headset back — including that $3,500 price tag. “It’s expensive,” Drummond acknowledges, adding, “I still think it’s a steal given how much amazing technology you’ve got, just those displays alone.”

However, the high price of the device has also held back adoption, with IDC estimating that the company shipped just 45,000 units during last year’s holiday quarter. “It doesn’t have great market penetration,” Drummond says.

Even people who own one may be reluctant to put it on every time someone sends them a two-minute clip or experience. “You don’t always have it to hand, and [you’ve got] the setup, teardown costs,” Drummond says. “We have phones with us all the time.”

It’s no secret that the Vision Pro was just a first step for Apple, which is reportedly working on its own smart and AR glasses. However, Drummond believes that such glasses products won’t be able to replace smart phones for years to come due to physical constraints. Phones are now capable of running complex machine learning models locally, while glasses will have much more limited power and compute capacities. Because of those constraints, they’ll likely function as another display for notifications and directions, similar to the Apple Watch. “I think the Watch and the spectacles from Apple will do much the same job,” he says.

This means that AR developers looking to build interactive entertainment experiences may find that the phone remains the best platform for the foreseeable future — a platform that is getting more capable by the day, thanks to AI.