A (0:04)

Welcome to the Tech We Write home for Friday, February 20, 2026. I'm Brian McCullough. Today, when your AI bot breaks your operations, Even when you're AWS, more on OpenAI's hardware plans and hey, remember perplexity. What's up with them is uber roadkill in the self driving car horse race and of course the weekend long range suggestions. Here's what you missed today in the world of tech. Sources say Amazon's AI tools have caused at least two AWS outages, including a 13 hour disruption in December after Kiro AI deleted and recreated an environment. Quoting the FT, Amazon Web Services experienced a 13 hour interruption to one system used by its customers in mid December after engineers allowed its Curo AI coding tools to make certain changes, according to four people familiar with the matter. The people said the agentic tool, which can take autonomous actions on behalf of users, determined that the best course of action was to delete and recreate the environment. Amazon posted an internal postmortem about the outage of the AWS system, which lets customers explore the cost of its services. Multiple Amazon employees told the FT that this was the second occasion in recent months in which one of the group's AI tools had been at the center of a service disruption. We've already seen at least two production outages in the past few months, said one senior AWS employee. The engineers let the AI agent resolve an issue without intervention. The outages were small but entirely foreseeable. AWS, which accounts for 60% of Amazon's operating profits, is seeking to build and deploy AI tools, including agents capable of taking actions independently based on human instructions. Like many big tech companies, it is seeking to sell this technology to outside customers. The incidents highlight the risk that these nascent AI tools can misbehave and cause disruptions. Amazon said it was a coincidence that AI tools were involved and that the same issue could occur with any developer tool or manual action. End quote. A U.S. grand jury has indicted two former Google engineers and one of their husbands for allegedly stealing trade secrets relating to the Tensor chip for Pixel phones. Quoting Bloomberg the three, all Iranian nationals, were charged in a U.S. indictment unsealed Thursday on 14 felony counts of conspiracy and theft of trade secrets and destroying evidence. Samena Gandali, 41, was a hardware engineer at Google in Silicon Valley, and her sister, Saror Gandali, 32, was an intern before they both joined another tech firm. Same husband, Mohammed Javed Khosravi, 40, applied multiple times at Google but was not hired and worked for a third tech company. The three were arrested Thursday and made initial appearances in federal court in San Jose, California, the Justice Department said in a statement. If convicted on the most serious charges, they could be sentenced to at least 20 years in prison, according to the statement. We have enhanced safeguards to protect our confidential information and immediately alerted law enforcement after discovering this incident, jose Castaneda, a Google spokesperson, said in a statement. Today's indictments are an important step towards accountability and we'll continue working to ensure our trade secrets remain secure. While working at Google, Semit Gandali sent more than 300 files, including company trade secrets, to a third party communications application based outside the U.S. according to the indictment. Saror Gandali sent 34 files, also including trade secrets, in the same way, the indictment says. Prosecutors alleged the three intended to provide the trade secrets to third parties, but it's not clear from the indictment if this happened. Google's internal security system detected the sisters downloading of its files and notified the FBI, according to the indictment. End quote. Sources tell the information that OpenAI now has more than 200 people working on a family of AI devices, including a smart speaker priced from $200 to $300, possibly smart glasses and a Smart lamp, Quote the smart speaker, the first device OpenAI will release, is likely to be priced between 200 and $300, according to two people with knowledge of it. The speaker will have a camera enabling it to take in information about its users and their surroundings, such as items on a nearby table or conversations people are having in the vicinity, according to one of the people. It people to buy things by identifying them with a facial recognition feature similar to Apple's face ID, the people said. While some OpenAI executives have suggested the company will tease its first device later this year, Peter Weilander, a vice president and general manager who's leading the devices team at OpenAI, wrote in a court filing earlier this month that the company doesn't expect the first device to ship to customers until next February at the earliest. Other devices, such as smart glasses, likely won't be ready for mass production until 2028, according to a person involved in AI glasses development. While the devices team has prepared prototypes for devices such as the Smart Lamp, it's unclear whether it will be released. The company's devices are still early, and details around their design and release schedule could change. A spokesperson from OpenAI declined to comment. During a presentation last summer, leaders from the device team told employees the device will be able to observe users through video and nudge them toward actions it believes will help them achieve their goals, said a person who attended the presentation. You could imagine the device observing its use, staying up late the night before a big meeting and suggesting that they go to bed, for example. Competition from other tech firms is putting pressure on the devices team, formed nine months ago after OpenAI acquired I O Products, the device startup started by CEO Sam Altman and former Apple design chief Johnny I've. That startup had been discussing potential devices since at least September 2023. Despite that deal, I've's involvement with OpenAI is complicated. He still runs his design firm, LoveFrom, as an entity independent of OpenAI, even though it is LoveFrom that is in char of coming up with potential OpenAI device designs. Meanwhile, OpenAI's internal devices team is in charge of making the hardware and the software powering it, as well as understanding how consumers will use the device. That division of responsibilities has sparked tensions. Some OpenAI staffers have complained that LoveFrom has been slow to revise its designs and shares little about its process of coming up with new ones. Even with others working on devices within OpenAI, to people with knowledge of the situation said that secrecy and meticulous focus on design is par for the course for Apple, where a number of device staffers and leaders came from. Apple has strict rules around which employees are allowed to know about various projects. In keeping with that approach, OpenAI's devices team itself is separate from the rest of OpenAI. While OpenAI's main office is in Mission Bay, the Devices team works out of a downtown San Francisco office in the Jackson Square neighborhood, not far from LoveFrom's office. I've makes the final call on most design choices, though he's only in the downtown San Francisco office about once a week or so. Despite that, the device's team still feels his presence very strongly, and staffers are known to refer to what they believe he would want frequently during conversations. End quote. Microsoft's AI Safety team has proposed technical standards for detecting AI generated content, but its CSO declined to commit to using that across its platforms, quoting MIT Technology Review to understand the gold standard that Microsoft is pushing, imagine you have a Rembrandt painting and you are trying to document its authenticity. You might describe its provenance with a detailed manifest of where the painting came from and all the times it changed hands. You might apply a watermark that would be invisible to humans but readable by a machine, and you could digitally scan the painting and generate a mathematical signature like a fingerprint based on the brush strokes. If you showed the piece at a museum. A skeptical visitor could then examine these proofs to verify that it's an original. All of these methods are already being used to varying degrees. In the effort to vet content online, Microsoft evaluated 60 different combinations of them, modeling how each setup would hold up under different failure scenarios, from metadata being stripped to content being slightly altered or deliberately manipulated. The team then mapped which combinations produce sound results that platforms can confidently show to people online, and which ones are so unreliable that they may cause more confusion than clarification. The company's chief scientific officer, Eric Horvitz, said the work was prompted by legislation like California's AI Transparency act, which will take effect in August, and the speed at which AI has developed to combine video and voice with striking fidelity. Nevertheless, Horvitz declined to commit to Microsoft using its own recommendation. Across its platforms, the company sits at the center of a giant AI content ecosystem. It runs Copilot, which can generate images and text. It operates Azure, the cloud service through which customers can access OpenAI and other major AI models. It owns LinkedIn, one of the world's largest professional platforms, and it holds a significant stake in OpenAI. But when asked about in house implementation, Horvitz said in a statement, product groups and leaders across the company were involved in this study to inform product roadmaps and infrastructure, and our engineering teams are taking action on the report's findings. End quote. Just because your team is small, that doesn't mean that they're not a target for bad actors. Cybercriminals know that lean teams often lack the resources to prevent or respond to a breach. The good news is even the smallest teams can foil cybercrime. 1Password provides simple security to help small teams manage the number one risk that bad actors exploit, which is weak passwords. Take the first step to better security by securing your team's credentials. Find out more at 1Passwordcomm/ride so you can secure every login that's 1Passwordcomm/ride. You know who we haven't heard from in a while? Perplexity quoting Wired Perplexity is abandoning plans to put ads in its AI search product as the industry looks for sustainable business models that won't hurt user trust. The change are part of a larger strategic shift for the company, which has long focused on disrupting Google's search business. Google is changing to be like Perplexity more than Perplexity is trying to take on Google, said a Perplexity executive at a press briefing on Tuesday. Executives spoke to the press on the condition of anonymity. Instead of chasing mass adoption, Perplexity will lean into its subscription business with a focus on becoming the most accurate AI service for developers, enterprises and consumers willing to pay a monthly fee. The company also plans to make partnerships with device makers a bigger part of its business moving forward. The move marks a major change for the company, which was one of the first AI firms to start experimenting with ads as far back as 2024. CEO Arvind Srinivas said on a podcast that year that he predicted ads would eventually be the company's core monetization engine. I think with advertising we could be really, really profitable, he added. Now executives say they're changing course because ads could make people mistrustful. Of Perplexity's responses, Anthropic offered a similar explanation for not putting ads in its chatbot clawed and poked fun at ChatGPT's ads in a Super bowl commercial earlier this month. But there may be other reasons for Perplexity not pursuing advertising early. Investors in Perplexity once believed the startup could reach hundreds of millions or even billions of users. But the startup's growth hasn't met expectations, according to a source close to the company. When the startup raised its Series B funding in 2024, board member and investor Kak Willem said in a blog post that Perplexity was capable of bringing the power of AI to billions. Two years later, that goal still seems a long way off. Data from the third party analytics firm SimilarWeb suggests perplexity had just over 60 million monthly active users across its website and mobile app in January. That's more than double the users Perplexity had last year, according to SimilarWeb. People also now access Perplexity via its AI powered browser comet, which SimilarWeb doesn't track. A source close to Perplexity says the agent in its comet browser reached 2.8 million weekly active users, which were also Perplexity subscribers in December 2025, down from a peak of 7.8 million weekly active users earlier in the without accounting for comment, Perplexity's user base on web and mobile is less than 10% of OpenAI's ChatGPT and Google's Gemini, which have 800 million weekly active users and 750 million monthly active users, respectively. One of the things that's starting to become clear to us is that Perplexity isn't for everyone, another Perplexity executive told the press. Advertising has been a strong business for companies like Google and Meta because they have hundreds of millions of free users. Without that scale, ads likely become a less appealing business model. Is there a horse race heating up in the self driving car space? Quoting the Journal Robotaxis are barely off the ground, but many investors already see the market as a two horse race, with Uber not among them. Because Tesla's service is designed to be available for anyone who owns a Tesla vehicle, with the company's full self driving feature enabled, it could scale up rather quickly, at least in theory. And while Waymo has partnered with Uber for service in the Atlantic and Austin markets, Uber's lack of involvement in subsequent market announcements has created a perception that Waymo is increasingly planning to go it alone. That has been a costly perception for Uber. The stock has lost nearly one quarter of its value over the past six months as Waymo has announced plans to expand into several new US Cities using its own app to run the service. The company has tried to address the fears, even going so far as to release a 13 page slide deck with its latest earnings report devoted to making the case for why autonomous vehicles aren't an existential risk to Uber. But that did little to soothe worries. Uber stock is down about 7% since the February 4th earnings report. It will take time for the market to fully digest the potential that Uber's relationship with Waymo, the most advanced AV operator in the market, is under severe strain, moffett Nathan analyst Mike Morton wrote in a note to clients following Uber's earnings. With the robotaxi market in its early stages, it seems way too early to call a winner that will rule the category. But Waymo's momentum is indeed undeniable. Uber is currently in the difficult position of having to prove that it won't be disrupted by a still burgeoning new technology. And that is going to take a while. Given the long time it will take for robotaxi services like Waymo and Tesla to scale, the debate on avs will take years, not months, to play out, said Morton of Moffett Nathanson. Only one long read for you this weekend. It's from the Verge. If you do take the plunge on a foldable phone, can you ditch your laptop? Quoting Allison Johnson There are still limitations to work with. Battery life is nowhere near as good as my MacBook, and when the phone is dead, so is my lifeline to the rest of the world. I found it's a good fit for shorter stints, maybe an hour or two, but I wouldn't try to make a full day out of purse computing without building in time to recharge. And you know what? I'm just fine with that. I don't need a foldable to be my all day laptop because that's not what purse computer is all about. I just want something easy to carry in a regular bag so I can be a human out in the world for a few hours. The friction of packing up my backpack is just enough to keep me from leaving the house most work days, but when it's just a matter of carrying the phone I was going to bring anyway, plus a little keyboard that fits into the small bag already on the back of my bike, suddenly it's a much easier trip to make. End quote. No bonus content for you this weekend. Talk to you on.