Latest AI News
Payroll startup Remote says it grew revenue 50% per employee without adding headcount
Remote, a seven-year-old, Amsterdam-based payroll service provider, says it recentlysurpassed $300 million in annual recurring revenueand became cash-flow positive. But the real story, it insists, is what happened behind the scenes: a 50% increase in revenue per employee after the startup adopted AI at every level of the organization. “As we are talking, on the second screen of my laptop, I have five different Claude instances running, building different things — and some of those are for me, but a lot of them are for Remote,” CEO Job van der Voort tells TechCrunch. This includes a Slack agent that summarizes discussions, as well as experiments with agentic AI; but the bigger picture is that Remote is now generating more revenue without increasing its headcount. According to van der Voort, the recipe behind these efficiency gains is AI adoption well beyond the CEO’s office or engineering department. Employees across all functions have been launching apps in Remote Labs, an internal marketplace built on the company’s own technology, and which shares similarities with the AI capabilities that the company is now opening up for its clients. Similarly to what Remote has been doing for its own processes, it is now helping clients create custom workflows. “We know that we’re ahead of most companies in that sense,” says van der Voort. “So we set up Remote Build, which is essentially what investors like to call ‘forward-deployed engineers’ — essentially people who work [directly] with our customers and prospects to do similar things inside of their organizations.” Van der Voort claims these gains could compound further. He says Remote’s core payroll business has grown more than 300% year over year — growth he attributes largely to AI adoption, though the company has not provided independent verification of that figure. Remote also says it now serves tens of thousands of companies navigating global employment compliance, a number that, like its ARR milestone, comes from the company itself. While Remote’s bread and butter is precisely this complexity, its staff also found relief in removing some of the repetitive and bureaucratic work required to pay workers in almost every country. “Obviously we’ve been automating a lot of that; that’s what we do,” says van der Voort. “But with AI that became easier, and arguably more fun than ever before.” Even though there’s nothing fun about payroll per se, van der Voort is also excited about the market opportunity it represents for his company. Despite its name — which might suggest a focus on distributed or remote workforces — he insisted the company targets all types of businesses, and the vast majority of its clients employ people in offices. “We do payroll for everybody, period.” Remote’s competitors largely went a different direction. Many went on to adopt an “all-in-one” HR platform model. But Remote sees the current AI wave and the subsequent commoditization of software as validation for its decision to stay focused on a hard problem. This also means that Remote has partners, and it is prepared to get out of the way to let them leverage AI. The recently launchedRemote MCPan interface based on the Model Context Protocol — a standard that lets AI agents securely interact with external software — grants AI agents and external platforms direct access to payroll and compliance data, allowing platforms like BambooHR and Workday to use Remote as an underlying engine. This goes hand in hand with the rise of agentic AI, which could see many companies virtually disappear — in a good way. “So if you use ChatGPT or Claude, you can control all of Remote; if you really wanted to, you don’t have to interact with our platform anymore,” van der Voort says. “I think that’s where the future goes.” According to van der Voort, the next step will be for AI agents to interact directly with Remote — with all the security standards required for an organization that deals with sensitive financial and personal information like payroll data. His own OpenClaw assistant — an open-source personal AI agent he named Jim — has served as an early explorer. “Jim can interact with Remote, and we build it in such a way that it is secure, so I don’t have to worry about my agent doing crazy stuff and messing things up. He has access to what he needs, but he cannot do destructive things. Those are the kinds of things that we’re really excited about, and it gives you a little bit of a taste of the future.” What’s happening internally at Remote may be another taste of the future. Like other tech companies,such as Spotify, it has embraced AI-powered coding, and the volume of contributions from its engineers has risen more than 60% over the last year. “And that’s accelerating, because if you look over the last month, more than 85% of all of our code is written by AI.” This has reduced Remote’s hiring plans, but hasn’t caused any job cuts, van der Voort says. He also noted that the company had not been planning a big recruitment campaign to begin with. “But certainly in some departments our plans were to hire more people than we did. [… ] What we’re doing now very actively is evaluating: ‘Do we actually need more people, or do we want to spend more time on upskilling the people that we have to use AI tools, and directly spending more money on AI?’.” His role is to “make sure that the company doesn’t run out of money and grows as fast as possible,” but rising AI costs aren’t a concern for him. “Our spend on AI is increasing, but we keep track of it, so it’s something that we’re happy with; and because we become more efficient as a company, we have some space to spend that on AI and those initiatives.” Remote’s trajectory offers one of the cleaner data points yet in the broader conversation about AI’s real business impact. The company isn’t just using AI to move faster — it’s using it to restructure how it scales. More revenue per employee, deferred hiring, and an expanding product surface area without proportional headcount growth is the operating model many companies are chasing. Another reason why van der Voort is happy with AI is that it has improved his own role. “This adds a whole new fun angle, I would say.”
View
In more good news for Amazon, Snowflake signs $6B deal with AWS for AI CPU chips
Cloud data storage giant Snowflake has signed a new $6 billion five-year agreement with Amazon Web Services, the companiesannouncedon Wednesday. Snowflake has always run on AWS, though obviously, these days, it is also available on Microsoft Azure and Google Cloud. For comparison on just how big this deal is for these companies, Snowflake has sold $7 billion worth of its services via AWS Marketplace total since it was founded in 2012, AWS says. So this new contract is close to all the money it has ever brought in from that cloud. It can do that because Snowflake’s customers are accelerating their spending on AWS as of late, Snowflake says, doubling in 2025 to $2 billion for that calendar year alone. What’s driving the growth is, naturally, AI. Snowflake has been offering its AI building tool, Cortex AI, for a couple of years now. It’s a tool that makes sense: Snowflake is where much of an enterprise’s data lives. The AI tool can provide features like a text interface for database queries (just ask, in regular language), summary reports, and so on. Of particular note is that Snowflake is signing this contract for more access to AWS’s home-grown ARM-based CPU chip, Graviton. As AI moves from training to daily usage to automation via agents, CPU usage skyrockets. While GPUs handle training and reasoning, CPUs handle most of the rest of the tasks associated with AI, particularly agents. Amazon CEO Andy Jassylast month boasted that Amazon’s own homegrown AI chipsoffer “better price-performance” than Nvidia’s offerings, though AWS still uses Nvidia’s chips in its cloud. Demand is so high for AI processing that cloud providers like AWS are deploying chips as fast as they can. On top of that, all of the major AI model makers (and many other AI offerings) have architected their apps specifically for Nvidia’s chips. Still, Amazon’s own chips are a more affordable option for the cloud giant to deploy. Amazon, ever the price-conscious company, says it passes those savings along to its customers. Consequently, these chips are luring in new multi-billion-dollar deals. Last month, for instance, AWS signed a deal to providemillions of Graviton chips to Metafor its growing AI compute needs. That was a big win for AWS because Meta had signed a $10 billion deal with Google Cloud a few months earlier. More than that, these deals are serving as notice to Nvidia that competitive CPUs from the cloud giants are attempting to come for its lunch. Google has also been making its own AI chips for years. Microsoft just launched itsMaia AI chip in January. Not surprisingly, Nvidia CEO Jensen Huang said last week that he’s more than ready to defend, and even grow, his turf. The new AI-specific CPU his company launched, called Vera,represents a ‘brand new” $200 billion market for Nvidia, he proclaimed after delivering another record-breaking quarter last week. And he’s already sold $20 billion worth, he said. While Nvidia may not be giving up market share to Amazon or any cloud provider that easily, AWS’s multi-billion-dollar cloud deals show how AI is lifting its boat. Whichever companies benefit most from the rise of AI in our work and home lives, the cloud providers are getting their share.
View
China is increasingly keeping its best AI talent to itself
For China’s top AI researchers, the borders are quickly closing. Researchers, startup founders and executives at private firms are nowreportedlysubject to travel restrictions, with some of the industry’s most prominent figures required to seek government approval before heading abroad. The restrictions reflect a wider shift in how Beijing manages the brain-drain in the AI sector, which has seenskyrocketing demand for talentto train and tweak AI models as the global tech industry taps into this new avenue to seek growth. In March 2025, the Wall Street Journalreportedthat Chinese authorities had been advising top AI founders and researchers to avoid traveling to the U.S., an early signal of just how closely Beijing has come to guard AI as both an economic asset and anational security priority. Restrictions appear to have intensified in the wake of Beijing narrowing its focus onthe Manus-Meta deal. China has barred Manus’ two co-founders from leaving the country while its regulators investigate whether Meta’s $2 billion acquisition of the AI startup runs afoul of Beijing’s foreign investment rules, according toThe Financial Times. The co-founders of Manus are now said to beexploring optionsto fulfill Beijing’s demand to unwind the deal, including raising about $1 billion from external investors to buy back the company from the social media giant. The AI race between the East and the West is closer than it’s ever been.Stanford’s latest indexshows the performance gap between the top U.S. and Chinese models had shrunk to just 2.7% as of March 2026, from about31%in 2023, raising fresh questions about how long America can hold its lead. The U.S. still dominates in terms of model quality and high-impact patents, but China is fast catching up if not outpacing American AI labs, in publications, citations and patent volume. In addition to travel restrictions, China reportedly plans to keep a check on U.S. capital flowing into its top AI firms, requiring government sign-off before tech companies like Moonshot AI, StepFun, and ByteDance can accept American capital, perBloomberg reportedin April. The news of travel restrictions follows a series of escalating economic countermeasures: in 2025, Beijing imposedtwo rounds of export controls on 14 rare earth materialscritical to high-tech military manufacturing, and separately barredstate-funded data centersfrom deploying foreign AI chips.
View
TechCrunch Disrupt 2026 Early Bird ticket savings end in 3 days
There are only 3 days left to save up to $410 on your ticket toTechCrunch Disrupt 2026. Early Bird pricing ends May 29 at 11:59 p.m. PT, and once the deadline passes, ticket prices increase. If you plan to attend one of the most influential gatherings in tech this year,now is the time to lock in your passbefore rates go up again. From October 13–15 at Moscone West in San Francisco, TechCrunch Disrupt brings together 10,000+ founders, investors, operators, and innovators driving the future of technology. Whether you’re raising capital, scouting investments, hiring talent, launching a startup, or building strategic partnerships, Disrupt is designed to put you in the middle of the conversations shaping what’s next. Here’s what you’ll gain by attending: Founder Pass: Connect with investors, gain practical insights, and access the tools and relationships that help startups grow faster. Investor Pass: Meet emerging startups, discover new investment opportunities, and maximize every conversation with curated networking tools. The countdown is on. Early Bird pricing disappears May 29 at 11:59 p.m. PT.Secure your ticket nowand save up to $410 before rates increase.
View
SOND, a sleep tech startup from Bose’s former head of sleep, exits stealth with $7M
Traditionally, sleep earbuds have been designed to mask outside noise and promote sleep with calming sounds. But today, a Boston-based startup calledSONDis introducing a new type of earbuds designed to actively intervene to encourage better sleep. Founded by apairof MIT grads, one who is Bose’s former Head of Global Sleep, SOND emerged from stealth on Wednesday with $7 million in funding. Together with the funding, the company introduced its debut product: Dreambuds, a closed-loop, in-ear system that captures 12 physiological signals from the wearer, then acts on them in real-time to help consumers get better sleep. Its initial investment of the $7 million comes from E14 Fund (an MIT-affiliated fund), Crosslink Capital, Ubiquity Ventures, Alumni Ventures, Meach Cove Capital, and Boston Scientific co-founder, John Abele. To work, the device tracks signals like respiration, heart rate variability, cardiorespiratory coupling, sleep staging, body position, snoring, and seismocardiography (SCG, or the mechanical vibrations of the chest wall produced by the beating heart). This sensor data streams in real-time to a cloud-based AI sleep coach that then selects a sleep audio program, or generates one on demand, learning over time which ones work best for the individual user. Users can also interact with the AI sleep coach directly by speaking, asking for sleep insights, or for specific sleep programs from SOND’s proprietary library of over 500 audio programs. (Users can also opt to stream podcasts through the case, if they prefer.) The AI coach can also generate audio, like a sleep story with a certain theme, when asked. Notably, the startup was co-founded and is led by CEOYadid Ayzenberg, who previously worked at Bose as its Head of Sleep Products, where he launched Bose’s Sleepbuds 2 and ran the company’s portfolio of other sleep products. When Bose decided to strategically exit the sleep business, Ayzenberg realized it presented an opportunity to form a startup dedicated to new products in this space, which led him to found SOND in February 2022. “I had spent, at this time, a significant amount of time around physiology, around sensors, around audio…I was meant to do this,” Ayzenberg told TechCrunch, while sitting at an outdoor cafe alongside co-founder and CTO Amir Lazarovich, formerly a senior software engineering manager at Google, alongside their prototype Dreambuds device. The co-founders met at MIT, a meeting that also had to do with sleep. Lazarovich, who was studying distributed systems, had just moved into a family dorm and didn’t have a mattress; Ayzenberg offered him one from his room to use instead. That chance encounter some fourteen years ago led to a lifelong friendship. After MIT, Ayzenberg founded a startup called The Sync Project, which mapped music to physiological factors like heart rate and heart rate variability. The startup was acquired by Bose after four years, and ultimately led to his work with the second generation of Sleepbuds. Bose customers often wanted more from their Sleepbuds than noise cancellation, Ayzenberg says: they also wanted sensors to track their sleep and help them improve it. At the time, technology was not quite at the point of being able to bundle a lot of sensors into a small, AirPods-like form factor while still conserving the device’s battery, however. But by the time Bose was exiting sleep wearables, that had changed. However, Ayzenberg cautioned, the Dreambuds shouldn’t be thought of as what could have been Bose’s Dreambuds III. Instead, he admits the earbuds from competitor Ozlo are more likely what would have been the next step. “We did something entirely different. Maybe the form factor is an earbud, but that’s where it ends,” he said. The system itself runs end-to-end without requiring a phone. Instead, Dreambuds’ charging case included Wi-Fi, Bluetooth, an OLED display, physical buttons, and a speaker. The latter will help you to wake up via your alarm even if you fall asleep before putting in the earbuds. The goal is to stop users from needing to pick up their phone to control the sleep tech’s system. “We have a running joke — we say giving an insomniac a phone is like running an AA meeting in a liquor store,” Ayzenberg says, with a laugh. “The idea here is that all you do is take the buds out and they’ll resume your sleep plan,” he explains. “You can also switch to other sleep plans. And you can talk to the coach, just double-tap and say, ‘I’m having trouble sleeping. I want this, or I want that.’” The sleep coach can help with particular sleep problems by referring to its data about what’s worked for you in the past, whether that was a breathing exercise, a calming track, a soundscape, binaural beats, or something else. Ayzenberg confirms the AI coach will never talk to you unless you engage it with the double-tap gesture, as he acknowledges that it could otherwise startle users or even creep them out. Lazarovich adds that the AI coach will respond based on the user’s current context. “For example, if you engage right before bedtime, it would ask you, ‘Are you ready to wind down?’ But if you engage after you woke up, it would ask you ‘How was your night?’,” he says. In addition to hearing your results from the AI coach, Dreambuds owners can review their data and hypnograms (sleep cycle graphs) in the companion app to learn more about their sleep patterns. The buds themselves have a unique look, as the team put the sensors facing out — opting for an artistic pattern of sensors instead of trying to hide the technology. The buds also feature wide-frequency drivers for high-fidelity audio, along with microphones and sensors for motion detection. SOND has run a couple of comfort studies and betas, and now aims to bring the devices into mass production by Q2 2026, following a crowdfunding campaign to raise additional funds. The company is currently accepting reservations on itswebsite.
View
ElevenLabs’ new music-generation model can switch genres mid-track
Voice AI companyElevenLabslauncheda new version of its music-generation model, called Music v2, that can switch genres mid-track. The company said that the model is designed to handle both complexity in vocals and composition. The new release comes nearly 10 months after the startuplaunched the first version of its music-generation model. ElevenLabs noted that the model can go from opera to heavy metal and back, deliver fast rap without losing coherence, and add non-musical sound effects to a track. With the new model, artists can pick a part of a song and re-create it using prompts without touching other parts of the track. Plus, instead of generating short clips, artists can build a song by sections, including the intro, verse, and chorus, then stitch them together. ElevenLabs added that the model performs more reliably across languages, lyrics, vocals, and arrangements. In the last few months, AI labs have been racing to release models that can generate professional-grade music.Google,Stability AI, andSunohave also released new music-generation models with capabilities to generate longer and more complex tracks. At the Google I/O developer conference, Google added the ability to easily create covers, edit songs by sections, and generate music videosusing its Flow Musictool. ElevenLabs emphasized that the new model is built on licensed data and cleared for commercial use, so users can freely use the tracks.Strikingdealswith labels is key, given that other AI music startups, likeSunoandUdio, faced court cases over copyright issues. The new model is available on ElevenLabs’ ElevenCreative tool for marketing and branding teams, along with its newly launchedElevenMusic platform for creating AI-generated songs, with availability on ElevenAPI coming soon.
View
Startup Battlefield 200 applications close today: Nominate a founder or submit your startup
Today is the final day to apply or nominate a startup forStartup Battlefield 200. Once the clock strikes 11:59 p.m. PT, the window closes on your chance to compete for $100,000 in equity-free funding, gain global visibility, connect directly with investors, and launch on theTechCrunch Disrupt stage. If you’re building a breakout startup — or know a founder who is — this is the moment to move. Apply nowfor the opportunity to join 200 of the world’s most promising early-stage startups at TechCrunch Disrupt. Founders, this is it. The application window closes tonight. The strongest startups are already in the arena, and applications always surge in the final hours. If your company has been nominated but you haven’t completed your application yet, don’t risk missing your shot by waiting until the last minute. And if you know a startup that deserves investor attention, media exposure, and a global stage,nominate them nowwhile there’s still time to apply before the deadline. Some of the most influential companies in tech history didn’t begin with perfect pitches or massive funding rounds. They started by taking a chance. Dropbox demoed to skeptics before cloud storage was mainstream. Cloudflare pitched before most people understood edge infrastructure. Discord entered as a scrappy gaming startup called Hammer & Chisel. All of them came through Startup Battlefield. That’s because Startup Battlefield 200 has never been about rewarding the most polished companies. It’s about identifying the most promising ones. Pre-launch is fine. Early traction is fine. No revenue is fine. What matters is whether you’re building something that genuinely changes an industry. The application itself is your first pitch. And today is your final opportunity to make it. Startup Battlefield 200 is where breakout startups get discovered. Selected companies will showcase at TechCrunch Disrupt in front of 10,000+ attendees, leading venture capital firms, global media, and the broader TechCrunch audience. Founders gain direct investor access, live exposure, and the opportunity to prove they belong among the next generation of category-defining companies. Every selected startup receives: Every selected company pitches live, whether on the Disrupt Stage or the Pitch Showcase Stage. Both put founders directly in front of investors, media, customers, and partners looking for what’s next. You do not need to win the competition for this experience to change your company’s trajectory. The exposure you get from this alone could be what pushes the needle. More than 1,700 startups have participated in Startup Battlefield over the years. Together, they’ve raised over $32 billion and produced more than 250 exits, including acquisitions by Microsoft, Google, Salesforce, Uber, and Amazon. Alumni include companies like Dropbox, Cloudflare, Discord, Fitbit, Mint, and Trello. Behind every one of those success stories was a founder willing to put their company forward before the rest of the world caught up. We’re looking for ambitious early-stage startups building innovative, potentially category-defining products. Applications are open globally across industries. Most selected startups are pre-Series A, though select Series A companies may qualify. To apply, startups should have: Thousands apply every year. Only 200 are selected. Just 20 finalists pitch on the main Disrupt Stage. One startup wins $100,000 in equity-free funding. The founders who change industries rarely wait until they feel completely ready. They apply before certainty exists. If you’ve been debating whether to submit, this is your final chance. The deadline closes tonight at 11:59 p.m. PT. If you’re building something category-defining — or know a startup that deserves the spotlight —submit your nomination and complete your applicationbefore time runs out.
View
AI coding startup Cognition raises $1B at $25B pre-money valuation
Cognition, the makers of the autonomous AI software engineer named Devin, has raised more than $1 billion at a $25 billion pre-money valuation, the company announced on Wednesday. That’s a major leap from its $10.2 billion post-money valuation when it closed a$400 million funding roundjust eight months ago in September. The round was led by Lux Capital and General Catalyst, with existing investors pouring in, including Founders Fund, 8VC, and others. The round also included new investors Ribbit Capital, Atreides, and Layer Global. This is a giant vote of confidence from top-tier VCs that there will be room for independent AI software coding startups. Last year, all signs pointed to model makers swallowing this hot market themselves. Certainly Anthropic’s Claude Code, OpenAI’s Codex, and maybe even Google’s coding agent Jules (afterGoogle’s acqui-hire deal of Windsurflast year) have captured a lot of it. But Cognition,which acquired the remaining bits of Windsurf last year, says it counts big enterprises like Mercedes-Benz, NASA, Goldman Sachs, and Santander as customers. It also says it’s reached $492 million in annualized revenue run-rate as enterprise usage of Devin has grown 50% month-over-month for the past six months.
View
Robinhood now lets your AI agents trade stocks
As the tech industry rallies around AI agents, some companies are building capabilities to enable AI agents to make payments and trade stocks on users’ behalf. Stock trading app Robinhood is also moving in that direction: The company on Wednesday said it is launching support for AI agentic trading, as well as a new agentic credit card. Robinhood said users on its platform can now create a separate account for their AI agents and connect them to a dedicated wallet. While these agents would be able to read and analyze users’ portfolios to come up with trading strategies and suggest investments, they’ll only be able to access the pre-loaded balance in the dedicated wallet to place orders. Users will get notifications of all trades their AI agent makes, and will be able to monitor their activities within the Robinhood app itself. For some trades, agents will show a preview that users may have to approve before the order is executed. The company said it has also built in fraud detection protection, in which a team from Robinhood would review suspicious trades and help users resolve disputes. Robinhood says users can connect their AI agents to its Model Context Protocol (MCP) service to do things like analyze concentration risk and sector exposure, execute trades, or look through analyst notes to identify new investment opportunities across various sectors. The agentic trading feature is launching in beta and only allows stock trading right now. The company says it plans to add support for options, crypto, event contracts, futures, and prediction markets soon. Robinhood is also debuting a new virtual credit card meant to be used by AI agents. With this card, users can connect their AI agents to the company’s banking MCP server to enable them to make payments. The virtual card is currently only available to Robinhood Gold Card holders, who can link their account to this new card. Users can set monthly limits on this virtual card, and can choose if their AI agent should seek approval every time it makes a payment. The company said its Robinhood Platinum Card will also get support for a similar virtual an agentic card feature when it launches later this year. Robinhood has been ramping up its AI efforts for the past few years. The company acquiredAI-powered research platform Pluto in 2024, and last year added an AI assistant thatoffers investment advice. “We’ve heard a lot of demand from our customers to bring their own tools, LLMs, and agents, and connect them to Robinhood. That is why we are launching our new products,” Abhishek Fatehpuria, VP of product at Robinhood, told TechCrunch over a call. Robinhood is not alone in enabling AI agents to make payments, with major players likeStripe,Amazon,Google, and newer startups like Prava Pay building products that give AI agents the ability to buy stuff on users’ behalf.
View
Tech CEOs are apparently suffering from AI psychosis
There is a certain wildness in the tech industry these days that both mimics previous eras of large changes, like cloud computing (runaway costs in the early days), and is like nothing we’ve ever seen before (record revenues accompanied by mass layoffs). A theory doing the rounds attempts to explain the phenomenon: Tech executives, especially CEOs, are collectively suffering from delusions of grandeur thanks to AI. And at least one tech CEO has said so out loud: Box founder Aaron Levie. “CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI,” Leviewrote on X. CEOs “play with AI,” develop a prototype, or generate a contract, to use Levie’s examples, and then make the leap to believing agents can do the work. But these top-level executives aren’t the people who have to review code, discover bugs, and identify calls to hallucinated libraries before software is deployed. They aren’t responsible for training AI models on a company’s idiosyncratic contract terms, nor do they have to spend days combing through contracts to find sneaky terms, as Levie indicates. In other words, Levie’s theory posits, CEOs don’t really understand processes well enough to know what really can and can’t be automated. But that lack of knowledge doesn’t stop them from acting on their beliefs. It’s important to note that Levie is not an AI hater. Quite the opposite. He mostly posts AI positivity on X to his 2.7 million followers, writing blogs titled,“Headless software is the future”on how software built for AI agents is the way forward. He also puts his money where his mouth is, backing AI startups as an active angel investor. So what are CEOs to do instead? Levie advises CEOs to use AI “a ton” to really see what it can and can’t do, “and come out the other side with an appreciation for both the upside and the real work.” CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI.So when they play with AI, they see the happy path results, often not considering the next 10 or 20 things that have…https://t.co/ne5mvJ4Rgx I have enough faith in humanity to believe that there are CEOs out there attempting to do just that, but right now, they seem to be in the minority. In only the first five months of 2026, the tech industry has already had nearly as many layoffs as in all of 2025: 115,430 people have been fired from 152 tech companies so far in 2026, compared to 124,636 people let go by 275 companies in 2025, according to industry layoff trackerLayoffs.fyi. And the bulk of companies have pointed to AI as a reason for cutting these jobs. Many argue that the biggest tech companiesare AI washing, or crediting AI productivity gains in the past or future, when other business decisions and metrics are really driving the cuts. Still, some of these stories are surprising. Zeb Evans, the CEO of project management and productivity software startup ClickUp,proudly declaredon X that he had laid off almost a quarter of his employees — 22% — after rolling out about 3,000 AI agents to do internal work. Evans swore this wasn’t done to reduce costs. Instead, he wants a workforce composed of people who run AI agents and spend their days quickly reviewing the agents’ work. He believes this will create a “100x org,” as he calls it. While AI can be a very useful tool, the data on AI and productivity doesn’t support such assumptions. By miles. A meta analysis of other research published in October in UC Berkeley’sCalifornia Management Reviewfound “no robust relationship between AI adoption and aggregate productivity gain.” Research published in March by theNational Bureau of Economic Research did concludethat AI adoption improved productivity, but noted “a productivity paradox, in which perceived productivity gains are larger than measured productivity gains.” After creating thousands of agents to work on tasks,researchers at MITconcluded that agents just aren’t doing human-quality work yet in many cases. They predict at the current rate of LLM improvement, models will “be able to complete most text-related tasks with success rates of, on average, 80%–95% by 2029 at a minimally sufficient quality level.” In other words, AI is on track to perform at base competence on most tasks in about three years. These researchers believe agents will need another few years to outperform humans. Meanwhile, research published in theHarvard Business Reviewshowed that when everyone is using AI to produce more stuff, the bottleneck simply shifts to executives. Their work awaits the people that must authorize all the stuff everyone is producing. If everyone is empowered to act, thenfrom what OpenAI experienced last year, we can tell that things may get out of control. Are CEOs ready for that? If not, the most certain outcome of the ongoing CEO AI psychosis will simply be organizational chaos.
View
YouTube will now automatically label AI videos
As AI video models become more powerful, YouTube is no longer solely relying on creators to label their AI videos — it will now automatically label videos on their behalf. The companyannouncedon Wednesday that its internal systems will apply labels when it detects that “significant photorealistic AI” has been used. YouTube will also be making its AI labels more prominent, so they’re easier to spot across both long-form videos and YouTube Shorts. AI labels on the video platform have been in usefor over two years, after YouTubeupdated its AI policiesand rolled out a tool in Creator Studio that required creators to disclose their videos included AI content that could be mistaken for a real person, place, or event. Videos that obviously depicted some sort of animated or imaginative scenario — like a unicorn prancing through a fantastical world — did not have to be labeled. The company says its policy around AI labeling hasn’t changed, but it will take a more active role in policing the content on its platform. The move follows Google’s release ofGemini Omni, a new family of multimodal AI modelsat its Google I/O developer conference last week that can output high-quality videos that reflect an understanding of physics, culture, history, and science. Starting in May, YouTube will now use new internal signals to help identify AI-generated content and label it accordingly, the company says. This doesn’t mean that creators shouldn’t continue to disclose their use of AI, but if they neglect to do so, YouTube will label the video for them. While creators whose content was misidentified will be able to update the disclosure status in a YouTube video, they won’t be able to remove those labels if the content was created with YouTube’s own AI tools, like Veo or Dream Screen, the company says. Labels will also be permanently attached to videos when the content containsC2PAmetadata indicating it was fully AI-generated. (Recently, OpenAIcommitted to the C2PA standard, joining Nvidia, Kakao, and Eleven Labs.) The addition of automatic AI detection functionality comes shortly after theexpansion of YouTube’s AI deepfake detection, which now allows any adult to scan YouTube specifically for face matches, after initial tests withcelebs, public figures,politicians, and other creators. YouTube says it will also make its AI labels more consistent and prominent. Before, labels would appear in the expanded description, unless the video touched on more sensitive topics like health or news; if so, a prominent label would appear directly on the video itself. Now, the labels will appear directly below the video player above the description for long-form videos and overload directly on YouTube Shorts. The company said moving the labels will make them more obvious to people who come across photorealistic, AI-altered, or AI-generated content on the site. Meanwhile, for AI video that is only slightly altered, animated, or unrealistic — like the above-mentioned prancing unicorn — the label will appear in the expanded description only. Notably, YouTube says that AI labels won’t have an impact on how a video is recommended or its ability to monetize. In addition to its policing of AI content, the company has been investing in AI for things likeits interactive search feature,Ask YouTube, aplaylist generatorfor YouTube Music,AI video summaries, andothergenerativeAIcreationtools.
View
ClickHouse triples anualized revenue to $250M, charting a path toward an IPO
Database provider ClickHouse has crossed $250 million in annualized revenue run rate, tripling its business from last year, Yury Izrailevsky, co-founder and president of product and technology, told TechCrunch. Israilevsky expects the revenue figure to reach the high nine figures by the end of the year. ClickHouse was valuedat $15 billionin January following a $400 million Series D funding round led by Dragoneer Investment Group. The latest valuation implies a steep multiple of over 60 times annualized revenue. The fast revenue growth and premium valuation position the less-than-five-year-old company for an IPO within the next few years, according to Izrailevsky (pictured left). ClickHouse joins a small, butgrowing listof tech startups signaling plans to go public as the IPO window is expected to be flung wide open by SpaceX’s historic June debut, followed by highly anticipated listings from OpenAI and Anthropic later this year. Last fall, the startup hired Jimmy Sexton, who previously ran investor relations at Snowflake, one of ClickHouse’s main competitors, as chief financial officer. Bringing on a CFO is often viewed as a signal that a company is preparing for public markets. The company has already acquired six startups, including Langfuse, which helps developers track and evaluate AI agent performance. Izrailevsky indicated that ClickHouse plans to remain acquisitive, looking to scoop up “relatively young, but showing very promising technology” startups, typically open-source, that complement its core product suite. The technology behind ClickHouse was originally developed inside Russian search giant Yandex 17 years ago, but spun out as an independent startup in 2021. ClickHouse has over 4,000 customers, including Anthropic, Meta, Capital One, and Decagon. The startup’s open-source database is designed to process the massive datasets required by AI agents. ClickHouse generates revenue by selling managed cloud services. Izrailevsky claimed that this commercial offering ultimately costs clients less than self-managing the open-source version. It “is something that’s a little counterintuitive, but it also has been a big tailwind for us,” he said.
View