Latest AI News
Bret Taylor’s Sierra buys YC-backed AI startup Fragment
Sierra, the customer service agent startup founded byBret Taylor, announced on Thursday that it has acquired theYC-backed French startup Fragment, which helps businesses integrate AI into workflows. This is Sierra’s third public acquisition. It previously bought Japan-based enterprise AI solutions company Opera Tech (which it acquired in late March) and voice agent company Receptive AI (which it also announced it acquired in late March). Fragment co-founders Olivier Moindrot and Guillaume Genthial will be joining the Sierra team. In a blog post, Taylor and his co-founder, Clay Bavor, wrote that Moindrot and Genthial will bring “valuable strength” to Sierra’s “agent development efforts in France.” Terms of the deal were not announced. PitchBook estimates that Fragment raised around $2 million through its seed round. Taylor, who also serves as OpenAI’s chairman of the board, co-founded Sierra alongside Google alum Bavor after stepping down as co-CEO of Salesforce in early 2023. The startup claims Casper, Clear, and Brex as customers and has raised more than$630 million in funding to date, from investors including Sequoia and Benchmark, giving it a $10 billion valuation.
View
Don’t stop hiring humans — stop hiring the wrong humans, with Jaspar Carmichael-Jack, Artisan
Surviving the early days as an AI startup isn’t just about making the technology work — it’s about hiring the right people, avoiding costly mistakes, and standing out in a crowded market. This week on Build Mode, Isabelle Johannessen sits down with Jaspar Carmichael-Jack, founder and CEO of Artisan, a fast-growing AI startup building AI employees for sales. Best known for its viral “Stop Hiring Humans” campaign, Artisan is rethinking outbound sales with AI, while still betting on hiring exceptional human talent. In this episode, they break down what it really takes to build and scale a venture-backed AI company, from Y Combinator to rapid growth. This conversation covers: Subscribe to Build Mode onApple Podcasts,Spotify, orwherever you like to listen. And watch the full videos onYouTube. Apply to Startup Battlefield:We are looking for early-stage companies that have an MVP. So nominate a founder (or yourself). Be sure to say you heard about Startup Battlefield from the Build Mode podcast.Apply here. TechCrunch Disrupt 2026:We’re back for TechCrunch Disrupt on October 13 to 15 in San Francisco, where the Startup Battlefield 200 takes the stage. So if you want to cheer them on, or just network with thousands of founders, VCs, and tech enthusiasts, thengrab your tickets. Use codebuildmode15 for 15% off any ticket type. New episodes ofBuild Modedrop every Thursday. Hosted by Isabelle Johannessen. Produced and edited by Maggie Nye. Audience development led by Morgan Little. Special thanks to the Foundry and Cheddar video teams.
View
Why Artisan still ‘loves hiring humans’ despite ‘Stop Hiring Humans’ billboards
Loading the player… Surviving the early days as an AI startup isn’t just about making the technology work — it’s about hiring the right people, avoiding costly mistakes, and standing out in a crowded market. This week on Build Mode, Isabelle Johannessen sits down with Jaspar Carmichael-Jack, founder and CEO of Artisan, a fast-growing AI startup building AI employees for sales. Best known for its viral “Stop Hiring Humans” campaign, Artisan is rethinking outbound sales with AI, while still betting on hiring exceptional human talent. In this episode, they break down what it really takes to build and scale a venture-backed AI company, from Y Combinator to rapid growth. This conversation covers: This conversation gets to the heart of building a startup: making the right hires early or paying for it later. Subscribe to Build Mode onApple Podcasts,Spotify, orwherever you like to listen. And watch the full videos onYouTube. Apply to Startup Battlefield:We are looking for early-stage companies that have an MVP. So nominate a founder (or yourself). Be sure to say you heard about Startup Battlefield from the Build Mode podcast.Apply here. TechCrunch Disrupt 2026:We’re back for TechCrunch Disrupt on October 13 to 15 in San Francisco, where the Startup Battlefield 200 takes the stage. So if you want to cheer them on, or just network with thousands of founders, VCs, and tech enthusiasts, thengrab your tickets. Use codebuildmode15 for 15% off any ticket type.
View
OpenAI releases GPT-5.5, bringing company one step closer to an AI ‘super app’
OpenAI on Thursdayreleased GPT-5.5, its newest AI model, which the company calls its “smartest and most intuitive to use model” yet. The algorithm comes with increased capabilities in a multitude of areas, with OpenAI co-founder and president Greg Brockman claiming that it also brings the company one step closer to the creation of OpenAI’s “super app.” On a call with journalists, Brockman said that the new model was a big advancement “towards more agentic and intuitive computing.” “This model is a real step forward towards the kind of computing that we expect in the future — but it is one step, and we expect to see many in the future,” Brockman said. “It’s a faster, sharper thinker for fewer tokens compared to something like 5.4. So this means that there’s just more frontier AI available for businesses and for consumers, which is part of our goal.” Brockman also said that the model was an additional step towardcreating a “super app”— a multi-purpose, Swiss Army knife of a program — which Brockman and co-founder Sam Altman have previously discussed launching. The co-foundersenvisioncombining ChatGPT, Codex, and AI browser into one unified service that can aid enterprise customers. Notably, the “super app” concept isalso a hot topicwith Altman rival (and former OpenAI colleague) Elon Musk, who has said he wants to turn X into its own so-called super app. OpenAI releasedits last modelonly last month, with a previous release in December and, before that, November. The company has continued to churn out new models at a crisp pace, a trend that company staff said should be expected to continue for the foreseeable future. “We see pretty significant improvements in the short term, extremely significant improvements in the medium term,” said Jakub Pachocki, OpenAI’s chief scientist. “In fact, I would say, like, I think the last two years have been surprisingly slow.” According to OpenAI, GPT-5.5 is designed to be useful across a broad array of categories, including foundational enterprise areas like agentic coding and knowledge work, but also in more experimental AI applications like mathematics and scientific research. The company also released data Thursday showing the model’s superior performance across a range of benchmarks. Compared to its previous models, and to models from competitors Google and Anthropic (like Gemini 3.1 Pro and Claude Opus 4.5), 5.5 consistently scores higher, according to OpenAI. OpenAI’s rivalry with Anthropic is never far from discussion, and GPT-5.5 offered another opportunity for the two companies to compare themselves to one another. One reporter during the press briefing asked if GPT 5.5. would have capabilities similar toMythos, the cybersecurity tool recently announced by Anthropic. (Mythos has experienced controversy in recent days due to a report ofunauthorized access to the program.) Mia Glaese, a member of OpenAI’s technical staff, said that GPT-5.5 would have a significant impact on the company’s approach to deploying its models toward digital defense. “We have a strong and longstanding strategy for our approach to cyber, and we’ve refined a durable approach to rolling out models safely,” Glaese said. Mark Chen, chief research officer at OpenAI, said that GPT-5.5 was better at navigating computer work than its predecessors, and also said that the model “shows meaningful gains on scientific and technical research workflows,” noting that the company feels it could really “help expert scientists make progress.” Chen also said it could assist with drug discovery, an area that has shownincreased industry interestover the last few years. GPT 5.5 is widely available starting Thursday, according to OpenAI. The company says that the model is deploying to Plus, Pro, Business, and Enterprise users in ChatGPT, while 5.5 Pro is headed to Pro, Business, and Enterprise users.
View
Meet Noscroll, an AI bot that does your doomscrolling for you
What if you could outsource your doomscrolling? That’s the premise behind the new startupNoscroll, which is offering an AI-powered bot that can browse your social feeds, news sites, and other online chatter, then text you when something important happens. “no feed. no brainrot. no ragebit,”readsNoscroll’s pitch to users. just signal.” X has the best information on the internet and the worst incentives & culture.meet noscroll — the AI that doomscrolls it for you and texts you just the things that matter.no feed. no brainrot. no ragebait. just signal.try it for free →https://t.co/XqdExWR13j🙅🏼♂️pic.twitter.com/EaHt2zfb7k The idea itself is fairly simple — it’s a bot that reads the web for you. But to work, there’s a lot that needs to go on under the hood. Nadav Hollander— previously the CTO at the NFT marketplace OpenSea afterselling his decentralized finance startupto the company in 2022 — said he built Noscroll because he found himself in a love/hate relationship with X. He was taking time off after leaving his job at OpenSea, and spent a lot of time on the social platform. “It’s phenomenally entertaining and really informative in ways you just don’t get from normal media,” Hollander told TechCrunch. “But it’s so toxic culturally, and it’s just very upsetting to read,” he said, comparing it to the nutritional equivalent of fast food. “You just feel terrible after it.” Hollander said he wanted to get off the app without missing out on the news and content. That inspired him to build Noscroll, which launched just a couple of days ago to the public. To get started with the service, you just text theNoscrollAI agent directly at (415) 583-7721, and it sends you a link to connect your X account to the service. This authentication provides Noscroll with information about your likes, bookmarks, and the accounts and posts you follow. The bot uses a variety of off-the-shelf AI models running on the company’s own proprietary infrastructure. The models have been customized with a lot of prompting, so the bot has its own unique voice and communication style. You can chat with the AI agent in natural language, telling it what sort of news or topics you want to keep up with, as well as what you don’t care about. It will then prepare a sample digest. To work, the AI pulls in information from beyond X, including news sites, blogs, Reddit, Hacker News, Substack, and more. It can even tap into things like research papers, local politics, or any other sources you may need. (You can recommend specific sources, too, if there’s something you want to make sure it checks.) Then, instead of spending your time scrolling through endless social media feeds to stay current on the news you care about, Noscroll will send you news digests via text at whatever cadence works best. For instance, a casual user might want to receive a weekly update on a topic, while a news junkie might want texts multiple times per day. These digests are essentially a collection of news links along with a brief AI summary of the article. If you want to know more, you can tap the links to open them up in your preferred web browser and read the article in full. You can also reply to the AI bot to ask questions and have conversations about the news you’re reading, too, as you could with other AI chatbots. Or you can add it to a group chat or Telegram group to have others engage with the service. (Other chat apps will be supported later on, we’re told.) The bot also knows when there’s breaking news worth seeing immediately and will text you as it’s happening. the sample batch sold it for me. subscribedhttps://t.co/W5ciDnQKRWpic.twitter.com/knq4qYmtBN — Alex Kwon (@startupoppa)April 22, 2026 Over time, the AI learns what you care about and uses that to better curate the types of information it sends you, the company claims. While the bot currently costs $9.99 per month to use, it will send you a sample news digest for free so you can customize it to your interests and try it out for 7 days. You can cancel the subscription at any time. Hollander notes Noscroll may experiment with variable pricing in the future. While there’s an obvious use case for those in the tech industry struggling to keep up with the constant stream of daily AI news and updates, Noscroll is not limited to tech topics. You can keep up with just about anything: reality TV, your favorite band, local news, your friends’ posts, your unread newsletters, or anything else that you find interesting. Hollander has been surprised to see how people are using it outside of tech. “People [are] following really niche anime industry news and local restaurant openings in Kyoto,” he says. Users are trying to stay on top of job listings, layoff tracking, and more. Journalists have also taken advantage of the tool to follow things like local politics and events. “I think the archetype that’s been interesting is anybody who has a professional need to be very online and follow things very closely. It’s quite useful to have a deputy who’s kind of doing that for you on whatever your beat is,” he adds. The AI bot has seen fast adoption, he says, and has already attracted investor interest. Hollander, who built the bot alongside his friend, an open source developer from the crypto world who only goes by the user name@z0ageon X, says the two have not yet made a decision on what to do with the inbound attention yet. Noscroll is available to try fromNoscroll.comby clicking the “text your agent” button.
View
Google Wants Your Employees to Manage Agents, and That May Rewire Work Itself
The Gemini Enterprise portfolio integrates tools that enable employees to manage AI agents rather than perform repetitive tasks.
View
AI galaxy hunters are adding to the global GPU crunch
NASA announced that it will launch the Nancy Grace Roman space telescope into orbit in September 2026, eight months ahead of schedule. The new space telescope is expected to deliver 20,000 terabytes of data to astronomers over the course of its life. That will add to 57 gigabytes of breath-taking imagery downlinked daily from the James Webb Space Telescope, which began its work in 2021, and the start of a survey later this year by the Vera C. Rubin Observatory in the mountains of Chile, which is expected to gather 20 terabytes of data each night. For comparison, the Hubble Space telescope, once the gold standard, delivers just 1 to 2 gigabytes of sensor readings each day. It’s been a while since all those readings were pored over by hand, but like everyone else with a pile of data, astronomers are now turning to GPUs to solve their problems. Brant Robertson, a UC Santa Cruz astrophysicist, has had a front-row seat to this step change in science while supporting or using data from these missions. Robertson has spent the past 15 years working with Nvidia to apply GPUs to the problems of understanding space, first through advanced simulations testing theories about supernnova explosions, and now developing the tools to analyze a torrent of data from the newest observatories. “There’s been this evolution [from] looking at a few objects, to doing CPU-based analyses on large scales of the data set, to then doing GPU-accelerated versions of those same analyses,” he told TechCrunch. Robertson and then-graduate student Ryan Hausen developed a deep learning model called Morpheus that can pore over large data sets and identify galaxies. Their early AI analysis of Webb data identified a surprising number of a specific type of disc galaxies and added a new wrinkle to theories about the development of our universe. Now Morpheus is changing with the times: Robertson is switching its architecture from convolutional neural networks to the transformers behind the rise of large language models. That will result in the model being able to analyze several times the area than it can currently, speeding up its work. Robertson is also working on generative AI models trained on space telescope data to improve the quality of observations collected by ground telescopes, which are distorted by Earth’s atmosphere. Despite advances in rocketry, it’s still hard to get an 8 meter mirror into orbit, so using software to improve Rubin’s observations is the next best thing. But he’s still feeling the pressure of global demand for GPU access. Robertson has used the National Science Foundation to build a GPU cluster at UC Santa Cruz, but it is becoming outdated even as more researchers want to apply compute-intensive techniques to their work. The Trump administration proposed cutting the NSF’s budget by 50% in its current budget request. “People want to do these AI, ML analyses, and GPUs are really the way to do that,” Robertson said. “You have to be entrepreneurial…especially when you’re working kind of at the edge of where the technology is. Universities are very risk averse because they just have constrained resources, so you have to go out and show them that, ‘look, this is where we’re going as a field.’”
View
Another customer of troubled startup Delve suffered a big security incident
The story of embattled compliance startup Delve keeps hitting twists and turns. TechCrunch has confirmed that Delve was the compliance company that performed the security certifications for Context AI, the AI agent training startup that last week disclosed a security incident whichled to a data breach at popular app and website hosting giant Vercel. On the other hand, Lovable, which had its own security incident, is no longer a Delve customer. To recap: Last month, Delve came under fire when an anonymous whistleblower alleged thatthe startup was faking customer dataand using rubber-stamping auditors in its compliance and certifications processes. Delve has denied those allegations. Soon afterwards, hackers attackedone of Delve’s security certification customers, LiteLLM, and planted malware in its open source code. After the incident, LiteLLM told TechCrunch it was dumping Delve and getting re-certified. Delve was alsoaccused of taking an open source tooland passing it off as its own work without proper license attribution. The startup’s reputation grew shaky, promptingY Combinator, where Delve graduated from, to sever ties. Fast-forward to last weekend, Vercel said hackers hadbreached its internal systems and accessed some customer data. The company said hackers broke in after an employee downloaded an app made by Context AI and connected that app to Vercel’s corporate account hosted by Google. The hackers abused that employee’s access to their Google account to break into some of Vercel’s internal systems. After Context AI was named in the Vercel attack, Gergely Orosz, author of the engineering newsletter The Pragmatic Engineer, saidin a post on Xthat Delve was the company that handled Context AI’s security certification. Context AI has now confirmed to TechCrunch that it did use Delve, but it has since ditched the startup and is in the process of getting re-certified. “Yes, Context was previously a Delve customer,” a spokesperson for Context AI told TechCrunch. “Following the reporting surrounding Delve in March, we transitioned our compliance program to Vanta and engaged Insight Assurance, an independent audit firm, to conduct new examinations. As part of the re-examination, we began updating our public materials, and we’ll share the new attestation when it is complete,” the spokesperson added. Security certifications on their own don’t stop security issues. They are intended to verify that a company has policies and processes in place to hinder attacks and reduce the likelihood of customer data being compromised. Case in point: Lovable was a Delve customer, butafter the whistleblower’s allegations came out,the vibe-coding platform said it had ditched the startup back in late 2025. The company has already re-completed one security certification, and is in process of redoing others, it said. Still, Lovable onMonday admittedthat it had inadvertently shared access to customer chat data publicly. The company also said it had dismissed vulnerability reports that alerted the company to the problem months earlier. Lovable apologized for initially denying there was a data breach, though it said the issue was caused by a configuration error, rather than a hack. There’s even weirder news swirling around Delve. The anonymous whistleblower, DeepDelver, haspublished another postalleging Delve was denying refunds to customers, but still took its team of more than 20 people to an offsite meeting in Hawaii between April 15 and April 19. The whistleblower shared some compelling receipts with TechCrunch that lend credence to the alleged Hawaii trip, but TechCrunch could not confirm other claims. After publication, Delve declined comment.
View
Grab a ticket today: The first StrictlyVC of 2026 kicks off in just a week in San Francisco
StrictlyVC’sfirst San Francisco event of the year is nearly here, and with just a week to go,now’s the time to grab yourself a ticket. The venture and founder communities will be gathering at the Sentro Filipino Cultural Center on April 30 for an evening packed with what StrictlyVC does best: engaging in enlightening conversations with excellent speakers, with ample opportunity for networking that drives outcomes. And we’re excited to announce one more speaker is joining our lineup:Eclipsefounder and CEOLior Susan, who has recently raised $1.3 billion to solely back “physical AI” startups. That extensive investment will be explored in detail in our discussion with him, as he shares his vision for what autonomy in the real world could look like as the convergence between the digital world of AI and our physical world evolves. Susan is well-versed in that space, having served in the military and leading Flex’s hardware investments. He’s seen firsthand the kinds of opportunities that lie in improving the industrial systems underpinning the modern world and using new technologies, supported by operators fueled by vision, capital, and innovative new technologies. And he’ll be interviewed by ourventure reporter Marina Temkin, so no follow-up opportunity will go unaddressed. Susan is joining the rest of our speakers for a particularly timely series of conversations.Replitco-founder and CEOAmjad Masadwill talk about the AI-driven shift in software development and what the next era of programming may look like, amid continued shifts in practices and employment within the industry. TDK VenturespresidentNicolas Sauvage, whose firm is also sponsoring the event, will joinTechCrunch editor-in-chief and StrictlyVC founder Connie Loizosfor a conversation about the ins and outs of corporate venture capital, early-stage investing, and key lessons founders must learn about raising from strategic backers. In addition,Campbell Brown, co-founder and CEO ofForum AI, will bring her perspective on building more trustworthy AI systems amid an increasingly skeptical environment around the accuracy of any information coming from any number of sources. All of which is informed by her tenure as Meta’s head of media partnerships and as a CNN host. And she’ll be interviewed by our newestsenior reporter, Tim Fernholz. And if you’ve never been to a StrictlyVC event before,this is a great one to start with. These meet-ups are a more intimate, informal affair that allow for candid and particularly instructive conversations that founders and VCs can use to get a head start on the competition immediately. You get the chance to actually connect with speakers, fellow investors, and founders, plus anyone else interested in learning more about the trends shaping Silicon Valley and beyond. Don’t wait to snag one of the few tickets still available. Block off your Thursday evening, and come to San Francisco on April 30 to join the StrictlyVC community.
View
Era raises $11M to build a software platform for AI gadgets
Earlier in April, the startupEraheld a gathering of artists who had received its developer kit in New York. The artists showed off the various mini gadgets they had built, like a souvenir that tells youfacts and jokes about France, a phone-like device that looks at your stocks and tells iftoday is the day you can quit your job, ora gadget that tells you about air quality. While all these devices are experimental, the common thread is Era’s platform, which allows hardware makers to create AI agents and orchestrations for AI devices. The company doesn’t want to create devices itself, but aims to enable others to do so by providing a software layer that could handle tasks like customized voice creation or adding intelligence to a classic device, such as headphones. The startup has raised $11 million in funding to date. This includes a $9 million seed round led by Abstract Ventures and BoxGroup, with participation from Collaborative Fund and Mozilla Ventures. Previously, the company had raised $2 million in pre-seed funding from Topology Ventures and Betaworks. Individual angel investors include Flickr co-founder Caterina Fake, iPhone keyboard creator Ken Kocienda, OAS founder Tony Wang, Little Guy co-founder Daniel Kuntz, Sandbar co-founder Mina Fahmi, ex Rabbit CPO ShaoBo Z, and Poetry Camera creator Kelin Zhang. Era was founded last year by CEO Liz Dorman, CTO Alex Ollman, and CPO Megan Gole. Dorman worked at Humane on AI orchestration andtransitioned to HP as part of the company’s acquisition. Ollman worked at HP on agentic frameworks for enterprises. Gole Sutter Hill Ventures on the Jony Ive and Sam Altman’s io project, and then later transitioned to Era. Era investor Casey Caruso, who is a founder and managing partner at Topology Ventures, said that the startup’s orchestration platform stands out because of its dynamic routing across models and managing real-world constraints like connectivity. Dorman said that the core idea behind Era was to build a platform that could power the next generation of devices, which might ditch the app model. “I think one of the incredible things that we can do with these AI models today is that you can replace that app layer. So what we’re building is the intelligence layer to allow anyone to create these types of intelligent objects, intelligent devices. And what we really believe is that the future of tech should not be made by people in San Francisco…It should not be people in their high fortresses who are so out of touch with reality, making devices and forcing them onto everyone. I want a choice over my devices again,” Dorman said. Currently, the company provides over 130 LLMs from more than 14 providers to enable different AI gadget form factors such as glasses, jewelry, and home speakers. Era thinks that as more form factors come to the forefront, hardware makers will need a software layer that can handle multimodal inputs and inference for them to power intelligent functions. “You can imagine this intelligence layer going to many different types of hardware. So we believe it’s not gonna be just glasses or rings or just bracelets. We’re gonna have a Cambrian explosion of what’s possible, and this is because tech is commoditized,” she said. Dorman noted that the startup’s platform is set up to scale across millions of devices. Plus, it can cater to custom AI device experiments that brands might do to appeal to certain users. The startup’s vision is that as more users adopt AI gadgets, it wants to enable users to choose their own memory and model providers in a privacy-preserving way. Just like it held a showcase with artists, it plans to make its platform available to the open source and maker community to show how its platform can power different kinds of devices. A big challenge in the AI hardware space is that there isn’t a model of a company that has found success. Humane was sold to HP, and Rabbit has been silent.Plaudhas found some success in meeting note-taking space, while startups likeSandbarandTayaare early. However, Era feels that as users see more use cases of AI devices, some will stick with them.
View
Muted Guidance Overshadows Infosys’ $20 Billion Revenue Milestone
Infosys reported $14.9 billion in large deal wins for the year, including $3.2 billion in Q4.
View
Infosys to Hire 20,000 Freshers in FY27
The hiring push comes despite a drop of over 8,000 employees in the March quarter.
View
