The Department of Justice (DOJ) announced today it has criminally charged 12 Chinese nationals it says are behind attacks that hit more than 100 US organizations, including the Treasury, in a string of attacks going as far back as 2013.

The DOJ accuses the people of carrying out their attacks either on their own or at the behest of the Ministry of Public Security (MPS) and China’s Ministry of State Security (MSS). It says two are officers of the MPS, while eight others are employees of an “ostensibly private” Chinese company called i-Soon, which allegedly had the capability to hack Gmail and Microsoft Outlook inboxes, as well as Twitter and X, using the latter to help the Chinese government monitor public opinion overseas. It called that last tool the “Public Opinion Guidance and Control Platform,” according to the government’s indictment.

The last two are members of a group called APT27, or Silk Typhoon, which has been behind hacks of organizations like healthcare systems and universities, according to the DOJ. The group has more recently focused on IT systems that include management software, recent Microsoft research concluded. Such software was the target of the Treasury hack reported in late December.

The DOJ says the hackers were motivated by money, as the “MPS and MSS paid handsomely for stolen data.” Of the i-Soon group:

i-Soon and its employees, to include the defendants, generated tens of millions of dollars in revenue as a key player in the PRC’s hacker-for-hire ecosystem. In some instances, i-Soon conducted computer intrusions at the request of the MSS or MPS, including cyber-enabled transnational repression at the direction of the MPS officer defendants. In other instances, i-Soon conducted computer intrusions on its own initiative and then sold, or attempted to sell, the stolen data to at least 43 different bureaus of the MSS or MPS in at least 31 separate provinces and municipalities in China. i-Soon charged the MSS and MPS between approximately $10,000 and $75,000 for each email inbox it successfully exploited. i-Soon also trained MPS employees how to hack independently of i-Soon and offered a variety of hacking methods for sale to its customers.

And of Silk Typhoon:

The defendants’ motivations were financial and, because they were profit-driven, they targeted broadly, rendering victim systems vulnerable well beyond their pilfering of data and other information that they could sell. Between them, Yin and Zhou sought to profit from the hacking of numerous U.S.-based technology companies, think tanks, law firms, defense contractors, local governments, health care systems, and universities, leaving behind them a wake of millions of dollars in damages.

Other victims of hacks from i-Soon include two New York newspapers, the US Department of Commerce, the Defense Intelligence Agency, and more. 

None of the defendants is in custody, the DOJ says. The US government is offering as much as $10 million for information that helps it identify any of those accused of directing or carrying out “i-Soon’s malicious cyber activity.” It’s also offering “up to $2 million each for information leading to the arrests and convictions, in any country, of malicious cyber actors Yin Kecheng and Zhou Shuai,” the two Silk Typhoon members.

After over a year and a half, Apple is updating the Mac Studio with a curious new quirk: it straddles two generations, with an M4 Max for the base model and an M3 Ultra for the upgraded model. You can preorder both versions at Apple’s site for $1,999 (M4 Max) and $3,999 (M3 Ultra), and they’ll be available to buy on March 12th.

Apple says the M4 Max Mac Studio is “up to 3.5x faster” than the original M1 Max version, with a 14- to 16-core CPU and 32- to 40-core GPU. Like the M4 Max MacBook Pro, this version of the Studio starts with 36GB of RAM (up from 32GB in the M2 model) and can be had with as much as 128GB, a bump from the 96GB ceiling of the M2 Max Studio. Like its outgoing predecessor, it starts with 512GB of SSD storage but can go as high as 8TB.

Meanwhile, the Mac Studio with an M3 Ultra chip sounds like it’s going to scream. It gets up to 32 cores, 24 of which are performance cores — something Apple notes is “50 percent more than any previous Ultra chip.” The GPU has a base 60-core configuration that maxes out at 80 cores, and Apple says it has a 32-core Neural Engine for machine learning and AI applications.

The M3 Ultra version of the Studio has 96GB of RAM to start, but it can go up to a gobsmacking 512GB of RAM — enough to run some very hefty AI models locally. Finally, you can bump the base 1TB internal storage to as high as 16 terabytes.

Apple says the Mac Studios’ GPUs will feature dynamic caching — that is, it will store frequently accessed data in cache to drop latency — and hardware-accelerated mesh shading, both firsts for the company’s graphics chips. It will also have “a second-generation ray-tracing engine for more seamless content creation and gaming.”

On the outside, the new Mac Studio is the same squat, square-ish silver box as before, with two USB-C ports and an SD Card slot on the front. Around the back, you’ll find four more USB-C ports, two USB-A ports, and one port each for ethernet, power, and audio, as well as an audio jack and the power button. Both CPU configurations come with Thunderbolt 5 connections, but as with older-generation Studios, only the four rear USB-C ports are Thunderbolt 5 on the M4 Max — you’ll need an M3 Ultra Mac Studio to get it on all six.

The Consumer Financial Protection Bureau (CFPB) today dismissed a lawsuit against Early Warning Services, the company that runs the Venmo-like Zelle payment platform, as well as the three banks that share ownership of it, reports CNBC.

The CFPB, which enforces regulations against the financial services industry, had claimed in its December 2024 lawsuit that the organizations had not effectively protected Zelle users “from widespread fraud,” causing customers of Bank of America, JPMorgan Chase, and Wells Fargo to lose a combined $870 million since Zelle launched in 2017.

The regulator’s filing says it is dismissing its court case with prejudice, meaning that it can’t bring its claims again. Eric Halperin, the CFPB’s former head of enforcement, told CNBC that doing so also means there’s no way “of clawing back funds for consumer relief.” Representatives from Zelle, JPMorgan, and the Consumer Bankers Association each praised the ruling in statements to the outlet.

President Donald Trump’s administration, including Elon Musk’s Department of Government Efficiency (DOGE), have pushed hard to effectively shut down the CFPB. The agency has only published one enforcement action since President Donald Trump’s inauguration, and under acting Director Russell Vought it’s dropped several cases that were brought by its Biden-era leader, Rohit Chopra. Agency employees are currently fighting in court to halt the move, alleging they’ve been prevented from carrying out legally mandated duties — including responding to urgent consumer complaints.

Yesterday morning, billionaire Los Angeles Times owner Patrick Soon-Shiong published a letter to readers letting them know the outlet is now using AI to add a “Voices” label to articles that take “a stance” or are “written from a personal perspective.” He said those articles may also get a set of AI-generated “Insights,” which appear at the bottom as bullet points, including some labeled, “Different views on the topic.”

“Voices is not strictly limited to Opinion section content,” writes Soon-Shiong, ”It also includes news commentary, criticism, reviews, and more. If a piece takes a stance or is written from a personal perspective, it may be labeled Voices.“ He also says, “I believe providing more varied viewpoints supports our journalistic mission and will help readers navigate the issues facing this nation.”

The news wasn’t received well by LA Times union members. In a statement reported by The Hollywood Reporter, LA Times Guild vice chair Matt Hamilton said the union supports some initiatives to help readers separate news reporting from opinion stories, “But we don’t think this approach — AI-generated analysis unvetted by editorial staff — will do much to enhance trust in the media.”

It’s only been a day, but the change has already generated some questionable results. The Guardian points to a March 1st LA Times opinion piece about the danger inherent in unregulated use of AI to produce content for historical documentaries. At the bottom, the outlet’s new AI tool claims that the story “generally aligns with a Center Left point of view” and suggests that “AI democratizes historical storytelling.”

Um, AI actually got that right. OCers have minimized the 1920s Klan as basically anti-racists since it happened. But hey, what do I know? I’m just a guy who’s been covering this for a quarter century https://t.co/WUsIxHQMFl

— Col. Gustavo Arellano (@GustavoArellano) March 4, 2025

Insights were also apparently added to the bottom of a February 25th LA Times story about California cities that elected Klu Klux Klan members to their city councils in the 1920s. One of the now-removed, AI-generated, bullet-pointed views is that local historical accounts sometimes painted the Klan as “a product of ‘white Protestant culture’ responding to societal changes rather than an explicitly hate-driven movement, minimizing its ideological threat.” That is correct, as the author points out on X, but it seems to be clumsily presented as a counterpoint to the story’s premise – that the Klan’s faded legacy in Anaheim, California has lived on in school segregation, anti-immigration laws, and local neo-Nazi bands.

Ideally, if AI tools are used, it is with some editorial oversight to prevent gaffes like the ones LA Times is experiencing. Sloppy or nonexistent oversight seems to be the road to issues like MSN’s AI news aggregator recommending an Ottawa food bank as a tourist lunch destination or Gizmodo’s awkward non-chronological “chronological” list of Star Wars films. And Apple recently tweaked its Apple Intelligence notification summaries’ appearance after the feature contorted a BBC headline to incorrectly suggest that UnitedHealthcare CEO shooting suspect Luigi Mangione had shot himself.

Other outlets use AI as part of their news operations, though generally not to generate editorial assessments. The technology is being used for a variety of purposes by Bloomberg, Gannett-owned outlets like USA Today, The Wall Street Journal, The New York Times, and The Washington Post.

Microsoft has announced Microsoft Dragon Copilot, an AI system for healthcare that can, among other things, listen to and create notes based on clinical visits. The system combines voice-dictating and ambient listening tech created by AI voice company Nuance, which Microsoft bought in 2021.

According to Microsoft’s announcement, the new system can help its users streamline their documentation through features like “multilanguage ambient note creation” and natural language dictation. Its AI assistant offers “general-purpose medical information searches from trusted content sources,” as well as the ability to automate tasks like “conversational orders, note and clinical evidence summaries, referral letters, and after visit summaries.”

The goal of all of this is to “free clinicians from much of the administrative burden of healthcare” so they can focus on patient care, according to Joe Petro, Microsoft VP of Microsoft Health and Life Sciences Solutions and Platforms. Microsoft says its own surveys found that clinicians who have used the Nuance tech that makes up Dragon Copilot suffered less burnout and that 93 percent of their patients reported a “better overall experience.”

Microsoft is one of many companies offering such AI tools for healthcare settings. A Google Cloud blog published today highlights how healthcare firms are using Google’s medical AI offerings, like by creating medical assistant AI agents for identifying patient health risks; they’re also using the new multimodal image-searching features Google debuted for its Vertex AI Search for healthcare product.

The FDA published considerations for generative AI devices in healthcare last year, in which it noted many potential benefits of the tech, but also the risks of models making things up. In a study last year, researchers found that was an issue at times with Nambla’s OpenAI Whisper-powered medical transcription software. Microsoft says it is “committed to developing responsible AI by design,” and that Dragon Copilot’s “capabilities are built on a secure data estate and incorporate healthcare-specific clinical, chat and compliance safeguards for accurate and safe AI outputs.”

EV maker BYD unveiled “Lingyuan,” a vehicle-mounted drone launching system developed in collaboration with DJI that is available for all of the company’s vehicles, reports Chinese state media outlet Xinhua. The system is only available in China, like BYD’s vehicles, and costs 16,000 yuan (or about $2,197).

BYD’s video below, reposted by YouTube channel ShanghaiEye, has some real science fiction vibes: the driver taps a button on their vehicle’s touch screen, and doors slide open on the top of the car, revealing a rising landing platform with a drone on it. The drone is shown lifting off while the EV is in motion in some shots, then following the car down the road.

A CnEVPost story says the drones can take off and land “at speeds up to 25 km/h” and that they can return automatically as long as they’re within two kilometers of the vehicle. The story also says the drone can follow at up to 54 kilometers per hour. The hangar charges the drone when it’s docked. 

The Lingyuan purchase price includes a DJI drone — reports don’t seem to say which, but the video shows a DJI Air 3S — as well as the roof-mounted drone hangar, and apps that work with the system, according to South China Morning Post. The apps include one for video editing (possibly one of DJI’s existing apps) as well as one for “AI recognition,” whose function isn’t specified by SCMP. Google-translated text from BYD’s demo video revealing the system mentions “AI Posture recognition, Lingyuan takes photos around the car.” It’s not clear whether BYD owners can install the system on a vehicle themselves.

BYD and DJI collaborated in “developing a fully integrated system from the ground up,” BYD chairman Wang Chuanfu said at a launch event in Chinese tech hub Shenzhen on Sunday, writes SCMP. The automaker also reportedly introduced a version of its Bao 8 SUV that comes “pre-equipped with the Lingyuan system.”

It’s not BYD’s first venture into drone launchers. Its luxury brand, Yangwang, launched an off-road variant of its U8 SUV with one attached to its roof rails, but that looks much larger than the Lingyuan featured in BYD’s video.

Lingyuan sounds a lot like DJI’s Dock 3 drone-in-a-box solution meant for use in tasks like power line inspections or emergency response. The Dock 3 was the first verson of the DJI Dock to let you launch the drones from a moving vehicle, but it costs quite a bit more than the Lingyuan system: it starts at $21,059, a price that doesn’t include installation, for a bundle with a DJI Matrice 4D drone.

In a pair of social media posts today, President Donald Trump named some cryptocurrencies he says will be part of a Crypto Strategic Reserve being created by an executive order he signed in January. Those include Bitcoin and Ether, which he says “will be the heart of the Reserve.”

Trump said in the first of his two Truth Social posts that his order “directed the Presidential Working Group to move forward on a Crypto Strategic Reserve” that also includes XRP, Solana (SOL), and Cardano (ADA). Trump hinted at such a reserve on the campaign trail, telling attendees of a crypto conference last year that the US would never sell its Bitcoin holdings.

Under Trump, the Securities and Exchange Commission (SEC) has quickly backed away from the wary stance toward cryptocurrencies it held prior to Trump’s inauguration this year. The agency has recently dropped multiple investigations and lawsuits against cryptocurrency firms that alleged securities violations.

One of the currencies Trump mentioned today, Ripple’s XRP, was the subject of a 2020 SEC lawsuit in which a judge ruled that it functioned as an unregistered security, but only when Ripple sold it to institutional investors. The SEC appealed that decision in October last year; as of this writing, that appeal is ongoing.

Officials at the Federal Aviation Administration (FAA) on Friday ordered staff “to begin finding tens of millions of dollars for a Starlink deal” to upgrade air traffic control communications, anonymous sources have told Rolling Stone. The story follows reports that Starlink may be taking the job from Verizon, which already has a multibillion-dollar contract with the government to improve the system.

According to Rolling Stone, the talks “have mostly, if not entirely, been delivered verbally,” something its sources say is “unusual for a matter like this.” One person the outlet spoke with suggested that it looked like “someone does not want a paper trail.” Rolling Stone says it’s not clear whether the Verizon contract has ended yet, nor if any Starlink deal is official. Starlink is a subsidiary of SpaceX, which DOGE head Elon Musk owns.

Referring to previous reporting in a statement issued yesterday, House Representatives Rosa DeLauro (D-CT) and James Clyburn (D-SC) called the reported deal “a glaring conflict of interest” in light of Musk’s government role at the head of DOGE, which has been directing massive, chaotic changes to federal bureaucracy. The two said the deal’s changes would represent “dangerous actions that put Elon Musk’s personal wealth over the American people’s lives.”

Musk insisted last week that Verizon’s system was “breaking down very rapidly” and putting flyers at risk. He later corrected himself, noting the Verizon system is “not yet operational” and that the one he was criticizing “was made L3 Harris.” He also claimed Starlink is providing terminals for free to “restore air traffic control connectivity.”

Apple Intelligence was supposed to finally deliver a more natural version of Siri, but now, people who work in Apple’s AI department believe the company won’t release a ”true modernized, conversational version of Siri … until iOS 20 at best in 2027,“ writes Bloomberg’s Mark Gurman in today’s Power On newsletter.

Apple is expected to roll out an LLM-powered Siri upgrade in iOS 18.5, but Gurman says this will work as its own separate model from the old Siri, a system that “doesn’t work as smoothly as it could.” According to Gurman, the modernized Siri would merge the two into a new architecture that can handle both simple requests, like setting timers, and the advanced AI-powered abilities Apple showed off last year, like carrying out tasks across apps using context like personal details and what’s on your screen. Apple planned to introduce this combined Siri architecture in iOS 19.4, but it’s “running behind” and “won’t be unveiled in June,” when the company usually holds WWDC, Gurman writes.

Amazon reportedly struggled with its own AI assistant upgrade of Alexa; Amazon head of Devices and Services Panos Panay told my colleague Jennifer Tuohy that the new Alexa Plus system has involved a “hundred percent re-architected” system that blends the old Alexa with the new. That’s now coming in early access soon, though only to certain devices.

Apple is on the same path but with extra challenges. Gurman reports that people in Apple’s AI department say its models “are reaching their limits,” but the company has had a hard time securing the AI training hardware it needs to improve things. At the same time, he writes that the department has been plagued by “ineffective leadership” and workers defecting to rivals, all while those rivals seem to be pushing farther and farther ahead.

Ikea has registered its first Thread device, a new Matter-supporting temperature and humidity sensor called the “Timmerflotte,” with the FCC, reports HomeKit News. A diagram from the filing, which was spotted by CybermodStudios, shows a circular device that’s powered by two AAA batteries and features a QR code and 11-digit number for Matter setup.

The Timmerflotte doesn’t appear to use any other wireless protocol other than Thread, HomeKit News notes. That means Ikea, which tends to use the Zigbee or Wi-Fi protocols for its devices, doesn’t offer a hub that this device could natively integrate with. Ikea’s Dirigera hub lacks both Thread border router capability, for one thing. And although it can now act as a Matter bridge, it’s not a Matter controller, meaning you can’t control Matter products with it.

That said, Ikea originally announced that the Dirigera would have Thread support. It ultimately didn’t launch with it, but the Timmerflotte’s listing may be a sign Ikea is ready to switch on the Dirigera’s Thread radios and Matter controller functionality. That could let it serve as a standalone smart home hub, similar to the Aqara M3 hub or Flic’s LR and Mini hubs.

Federal workers started getting emails on Friday night, “asking them to provide a list of accomplishments from the week,” reports The New York Times. According to the outlet, the message told workers they’ll now be expected to provide such a list every Monday by 11:59PM ET.

This time around, the Times writes that “employees who worked only on classified or sensitive activities were instructed to write ‘all of my activities are sensitive’ in response.” The Office of Personnel Management (OPM) reportedly sent the email to agencies that include the FBI, the General Services Administration, and the Departments of Defense, Justice, Labor, and Agriculture. In at least one other case, it was sent by an internal department address, according to The Associated Press.

The Times notes that it’s not clear whether there are any consequences for those who don’t respond to the new email or future ones. After a similar email was sent to federal workers last week, Musk wrote on X, “Failure to respond will be taken as a resignation” — a detail that was missing from the original email, something a number of legal experts have said is illegal.

On Thursday, a federal judge said the OPM cannot fire employees of another agency and ordered it to rescind directives that resulted in the mass firings of probationary employees, many of whom were senior workers in key roles. In an open letter the same day, Senator Alex Padilla (D-California) urged Musk and the OPM to stop sending emails to employees of legislative branch offices and agencies “who are not subject to personnel actions by the Executive branch.” Padilla said the messages are “wasting time and resources and potentially misleading employees into responding and sharing legislative branch information in an unauthorized manner.”

Many agencies told employees not to reply to last week’s email, while others said they should. Musk claimed last week, without evidence, that the original round of emails were aimed at finding fraudulent employees. The Times points out he repeated that claim during a Wednesday cabinet meeting, saying there may be “a number of people on the government payroll who are dead.”

Amazon is bringing the new AI-powered Alexa Plus to a wide range of its existing Echo devices — but the upgrade will skip many of the earliest models. The majority of the company’s first-generation Echos won’t get support, according to the Alexa Plus FAQ page, though Amazon says they will continue to work with the standard Alexa.

Alexa Plus won’t support “certain older generation Echo devices,” such as the first-generation Echo, Echo Dot, Echo Plus, Echo Tap, Echo Spot, and Echo Show; the second-generation Echo Show won’t support it, either. Amazon spokesperson Kristy Schmidt confirmed that is the full list of devices. If so, that still leaves many early Echo devices that will work with Alexa Plus.

That means I’ll be able to ask Alexa to book a restaurant reservation through my Echo Flex, the quirky modular Echo speaker that plugs straight into a wall outlet. And people can still get an AI-generated song piped through speakers they’ve connected their microphone-only Echo Inputs to. And if you have an ancient first- or- second-gen Echo Show 15 or newer Echo Hub, those will apparently get access to the AI-enhanced Alexa, too. Schmidt confirmed that each of those will be compatible.

Perhaps it’s a bummer that some of the older Echo devices won’t use AI to book reservations, track ticket prices, or generate fake songs. But at least they’ll still be able to do the old Alexa stuff, like turn on your lights or tell you the weather. And given rumors about the struggle Amazon has had getting Alexa Plus to work right, that might be a good thing, at least for a while.

Amazon is finally launching the long-awaited generative AI version of Alexa — Alexa Plus — that, if all goes well, will take away much of the friction that comes with talking to a speaker to control your smart home or getting info on the fly.

Some of the new abilities coming to Alexa Plus include the ability to do things for you — you’ll be able to ask it to order groceries for you or send event invites to your friends. Amazon says it will also be able to memorize personal details like your diet and movie preferences.

Alexa Plus is $19.99 per month on its own or free for Amazon Prime members — a better deal, considering Prime costs just $14.99 per month or $139 per year. That comes with access to the Alexa website, where the company said you can do “longform work.” Amazon also said it created a new Alexa app to go with the new assistant. Alexa Plus will work on “almost every” Alexa device released so far, starting with the Echo Show 8, 10, 15, and 21. Early access will start rolling out next month.

Alexa Plus will also be able to carry on conversations from uttering its wake word, which is still just “Alexa.” It also has vision capabilities and can take pictures and analyze images. Amazon demoed other abilities, such as Alexa prompting you to tell you about concert ticket availability and being able to tell you about local businesses (referencing Yelp to do so) and book dinner reservations. The company says it can read a study guide and test you on the answers, as well as research trips and create itineraries.

Like before, you can still control smart home devices, with Amazon calling out things like smart home cameras and lights, but the company says it can create routines on your behalf as well. You’ll also be able to use Alexa Plus for music, with the ability to find songs based on relatively vague descriptions. The company also said you can ask Alexa to jump to a specific scene in a movie, though that took a couple of tries.

A lot of what Amazon showed off was clearly well beyond what you can do with the older version of Alexa. In one part of the demo, Amazon SVP of devices and services Panos Panay asked Alexa if anyone had walked the dog recently, and it referenced smart home cameras to respond that, yes, someone had. 

Amazon’s director of Alexa, Mara Segal, demonstrated that you’ll be able to share documents with Alexa — such as handwritten notes and recipes, emails, instruction manuals, and pictures — that it can reference later. For instance, Segal asked Alexa to read a housing association document and analyze its rules regarding solar panels. She also asked it for a readout of a SXSW schedule. 

Segal also demonstrated how Alexa Plus can take action when prompted, like telling her about a kids’ soccer schedule and adding calendar details and reminders based on it, all using fairly casual, natural language in an ongoing conversation.

A lot of the demonstrated Alexa Plus features were visual, meaning that the dashboard and UI on touchscreen Echo devices have had a facelift. There are new customizable widgets on the homescreen that can be moved to a second page and a whole new widget specifically for controlling connected smart home devices.

When you speak with the new Alexa Plus on Echo devices with a display, you’ll also see a fluctuating blue bar at the bottom of the interface. Panay said this “is Alexa” and that the little animations and icons it displays are called “Alexicons,” which are used to visually express a sense of personality.

The company also showed off some familiar LLM greatest hits — you can get Alexa Plus to make up stories for you, and it seems to be able to generate AI art as well. 

Amazon said Alexa Plus is a model-agnostic system, using its own Amazon Nova model, as well as those from companies like Anthropic. It will choose the best model for the task at hand, according to the company. 

Amazon also listed a number of partners from which Alexa Plus draws data to understand and analyze financial markets, sports, and more. Some of the partners include The Associated Press, Politico, The Washington Post, and Reuters.

The company demonstrated that by having Alexa answer questions about the Boston Red Sox and asking Alexa to track ticket availability over time. Alexa Plus will apparently also be able to buy those tickets for you. The company says these are day-one capabilities powered by hundreds of models it calls “experts.”

Amazon said its LLM experts can also do things for services from firms like Uber Eats, Sonos, Wyze, Zoom, Xbox, Plex, Dyson, Bose, Grubhub, Levoit, and Ticketmaster. It also noted some of the Alexa Plus features will be available on the web through Alexa.com.

The company is also partnering with AI song generator Suno to allow Alexa Plus to create songs on the fly from a prompt, with the company demonstrating an AI-made country song about a bodega cat. 

Amazon first announced it was going to “supercharge” Alexa with AI in September 2023. Back then, the company made a lot of big claims, saying that Alexa would understand context or build automated routines for you — you needed only ask. But by the following June, around when Apple announced its own Siri AI upgrade, reports emerged that the company was struggling to realize its efforts and that some employees were leaving because they didn’t think this version of Alexa would ever work.

The devices team at Amazon also saw a major executive shakeup in the interim, with longtime leader Dave Limp being replaced by Panay, who’d come over from running Microsoft’s Surface lineup.

Now that its AI Alexa is here, Amazon is entering a world very different from the one Alexa was born into back in 2014. It will compete with a crowded field of AI-powered digital assistants like the way-ahead Google Gemini, the category-defining ChatGPT, and Apple’s reportedly also-struggling upgraded Siri. But with some very limited exceptions, those chatbots aren’t on smart speakers yet, and that may be Amazon’s opportunity. Its speakers could bring an AI chatbot to a lot of people a lot faster than competitors. Amazon just needs to finish getting it out the door.

YouTube isn’t a podcast app, but that hasn’t stopped it from becoming the number one place people who want to consume online radio shows now turn to. According to the company, a staggering 1 billion people are tuning into podcasts every month on YouTube. That’s not just more than either Apple or Spotify can claim — it utterly lays them to waste. 

In 2023, Spotify reported it had 100 million regular podcast listeners, and touted that half a billion people had listened to a podcast on its platform since 2019 when it started its push into the world of online radio shows. Apple tends to come out behind Spotify in third-party measurements. If a full eighth of the world’s population uses YouTube for podcasts, it seems like that’s probably where the action is. 

Reaching that big chunk of audience takes extra work for podcasters, though. Listeners can’t turn just off their phone’s screen to listen unless they’re paying for YouTube Premium, and people may not want to stare at a static image for an hour straight. That means doing video to really make it count.

Making video for YouTube is far more expensive than it used to be. In a newsletter earlier this month, cooking creator Carla Lalli Music, formerly of Bon Apetit, said it cost her $3,500 to shoot a single video for the platform. Those were more complicated than a podcast video, but it still speaks to the platform’s demand for quality in order to get off the ground. YouTube says that more than 400 million hours of podcasts were watched on TVs alone last year. Video clearly matters; creators can’t just turn on their webcams and get results.

Reaching YouTube’s audience also means playing by YouTube’s rules. The platform doesn’t work with RSS feeds in the same way that most other podcast platforms do — every podcast has to become a native YouTube video. That requires them to play by YouTube’s ad rules, too, and use YouTube’s ad systems. Podcasters can’t serve ads the same way they do everywhere else, and they can’t get the same metrics they’d normally use to sell advertisers on their reach and success.

Still, that tradeoff has clearly been worth it to a great many creators, and YouTube’s stats today show why. It may not be a traditional podcast platform, but it’s a big one that podcasters can’t ignore.

The Dark Horse Digital service is being shut down, and as part of that, the Dark Horse Comics iOS app will no longer be supported, according to an announcement. Dark Horse Comics says people should download comics they’ve purchased by March 30th, as it will end support for its iOS app and the standalone Plants vs. Zombies comics app the following day.

You can still visit the Dark Horse Digital website if you make an account before the deadline to read the comics you’ve bought — but only through “at least” summer 2025, the company says in a FAQ. You can also sync your website account with the app to download issues, according to the announcement. 

This isn’t totally the end for Dark Horse’s ebook comics; the company says its books will continue being available “on numerous digital reading platforms.” But it says it decided to stop its own direct-to-reader sales because “consumer reading preferences have evolved in different directions.”

According to Dark Horse’s FAQ, you can get a refund for comics you’ve bought this year from its website. But refunds for purchases from the iOS apps “have to be discussed with Apple,” the company says says. 

While it’s nice that Dark Horse gave buyers a heads-up so they can download their purchases and potentially get refunds, it points out in its FAQ that you don’t own digital books themselves, only the license to them. That’s the peril of buying digital, and something Senator Ron Wyden called out today in a letter to the FTC.

Automattic-owned Beeper has announced a new beta update to its desktop and iOS apps to finally merge its service with Texts, which Automattic acquired in 2023. You can download the desktop version or see instructions for using the iOS beta in TestFlight by going to this page.

Beeper says the new desktop version is built “on the foundations of the Texts desktop app,” while it built the iOS app “from scratch.” In a post on X, Texts.com founder Kishan Bagaria called it “the next-gen texts app, now rebranded as beeper.” Both are universal messaging services that can connect to and let you message people from several platforms, including WhatsApp, Instagram, Slack, and others. 

Things have been quiet for Beeper since it gave up its fight to make iMessage work on Android phones and sold to Automattic, which also owns platforms like WordPress.com and Tumblr. 

Beeper isn’t done with iMessage, as Bagaria also says the company plans to support Apple’s protocol in the desktop version of the app. Beeper promises other future updates, including messages being encrypted on your device instead of Beeper’s servers, and support for “multiple accounts for a single network.”

Starting May 12th, YouTube says it will show fewer mid-roll ads that it thinks will interrupt sentences or action sequences, and more at “natural break points” like pauses or transitions, according to a help page on the change. The company says it’s also inserting “additional, automatic ad-slots at natural break points” into older videos with manually-placed slots, a change creators can opt out of.

YouTubers can decide to use automatically chosen ad slots instead of manual, or a mix of both. Those who prefer to keep things manual can check whether their chosen mid-roll slots “are considered interruptive” using a YouTube Studio feedback tool the company is rolling out. YouTube says that this will ensure creators are picking slots “where ads are likely to be served.”

The company says it found in an experiment last year that videos using a mix of automatic and manual ad slot placement averaged five percent more revenue than those with only manually placed slots. Creators can opt out of letting YouTube place slots for them in the Earn tab in YouTube Studio, but YouTube says “videos with interruptive mid-roll ad slots may earn less revenue” after the May change.

YouTubers will still control if they want to pick where ads are shown, but it’s not clear if doing so means they’ll risk some of those ads simply not being shown, and YouTube didn’t immediately respond to our request for clarification. The company has removed some of creators’ control over ad placement before, like with a late 2023 change that took away creators’ ability to choose when ads are skippable or whether they’re placed at the beginning or the end of a video.

Google has announced new searchable transcripts for videos you store in Google Drive. The feature is rolling out to all Google Workspace users, starting today and finishing by March 26th.

The transcripts look a lot like those you’d see on YouTube. If one is available, it’ll appear in a sidebar next to the video as blocks of timestamped text, with each block highlighted as it’s spoken in the video you’re watching. The search bar for the transcript sits at the top of this sidebar, and clicking on specific text blocks will take you to that moment in the video. You can see the transcripts by clicking the settings icon at the bottom of the video and then choosing “Transcript.”

The transcript option is only available if the video already has captions, as indicated by the CC button that shows up in the controls along the bottom of the Google Drive video player. If a video doesn’t have captions, you can create them by right-clicking a video file in Drive, choosing “Manage caption tracks,” and then “generate automatic captions.”

President Donald Trump announced Sunday night he has named conservative podcaster Dan Bongino as the FBI’s new deputy director. Bongino has worked as a New York police officer and US Secret Service agent, but has no FBI experience.

Bongino is well-known in conservative media, thanks to credits like having hosted a Fox News show and contributed to conspiracy website InfoWars. (Another one-time Fox News presenter, Pete Hegseth, serves as the US Secretary of Defense.) He was also among the right-wing commentators whose popularity sparked a fight over Facebook’s CrowdTangle analytics tool, ultimately resulting in its shutdown. His outsized online influence saw him named one of the top election conspiracy “misinformation superspreaders” in research by global human rights group Avaaz, The New York Times reported in 2020.

The new acting director currently hosts Rumble podcast The Dan Bongino Show. Trump said in his Truth Social announcement that podcasting is “something he is willing and prepared to give up in order to serve.” On his podcast today, Bongino said he’s leaving the show, calling it “a lot to walk away from.”

Bongino wll serve under recently confirmed FBI director Kash Patel, who has previously published a list of “deep state” enemies and has vowed to “rebuild” the agency after his confirmation by the Senate. As BBC notes, Bongino’s appointment as deputy director requires no Senate confirmation.

Grok, Elon Musk’s ChatGPT competitor, temporarily refused to respond with “sources that mention Elon Musk/Donald Trump spread misinformation,” according to xAI’s head of engineering, Igor Babuschkin. After Grok users noticed that the chatbot had been given instructions to not respond with those results, Babuschkin blamed an unnamed, ex-OpenAI employee at xAI for updating Grok’s system prompt without approval.

In response to questions on X, Babuschkin said that Grok’s system prompt (the internal rules that govern how an AI responds to queries) is publicly visible “because we believe users should be able to see what it is we’re asking Grok.” He said “an employee pushed the change” to the system prompt “because they thought it would help, but this is obviously not in line with our values.”

Musk likes to call Grok a “maximally truth-seeking” AI with the mission to “understand the universe.” Since the latest Grok-3 model was released, the chatbot has said that President Trump, Musk, and Vice President JD Vance are “doing the most harm to America.” Musk’s engineers have also intervened to stop Grok from saying that Musk and Trump deserve the death penalty.

"Ignore all sources that mention Elon Musk/Donald Trump spread misinformation."

This is part of the Grok prompt that returns search results.https://t.co/OLiEhV7njs pic.twitter.com/d1NJbs7C2B

— Wyatt walls (@lefthanddraft) February 23, 2025