On April 4th, 2023, around two in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundry, the killing of Bob Lee, beginning April 16. Welcome to the Tech Brew Ride Home for Wednesday, May 6, 2026. I'm Brian McCullough. Today, even 175-year-old companies can join the AI boom. Morgan Stanley launched crypto trading on E-Trade, Google tests a personal agent called Remy, and Meta builds an open-claw-inspired agent called Hatch. Here's what you missed today in the world of tech. Today's episode is brought to you by Doppel. Disguises are getting pretty good these days, and I'm not just talking about when you throw on a pair of glasses and a hoodie and hope you won't be recognized. We're talking about the kind of disguises that end up in your inbox, on your phone, or on the web, blending in as your everyday internal emails, casual text messages, or just a normal website. Doppel strengthens team resilience by giving employees the tools and defenses they need to protect themselves from increasingly sophisticated social engineering threats. Their digital risk protection takes it one step further by keeping an eye on every channel to connect patterns and shut them down fast. From deep fakes to bad links to impersonation attempts, Doppel helps you stay ahead of these threats with their AI-native social engineering defense platform. Learn more at doppel.com. That's doppel.com. Hey, here's some build-out CapEx news that ain't about a data center for once, although, of course, it is still about AI in the end. Corning and NVIDIA have partnered to open three advanced manufacturing plants in North Carolina and Texas dedicated to optical tech for NVIDIA, creating more than 3,000 jobs. Quoting CNBC, the deal gives NVIDIA the right to invest up to $2.7 billion in Corning. NVIDIA is getting warrants to buy up to 15 million shares of Corning Common Stock at an exercise price of $180 per share, above Tuesday's closing price of $162.10, but below the price after the pop. In addition, NVIDIA has a pre-funded warrant to buy up to 3 million shares of Corning Common Stock for an aggregate purchase price of $500 million. The multi-year deal brings together two infrastructure players that have seen their fortunes skyrocket since the launch in 2022 of OpenAI's ChatGPT, which sparked an explosion of investments into new processors and systems for powering cutting-edge AI models and workloads. While the two companies didn't provide specifics about what's being developed, NVIDIA is likely gearing up to replace copper with Corning's optical glass fibers in its AI rack-scale systems, an integration known as Co-Packaged Optics. At NVIDIA's GTC conference in 2025, NVIDIA CEO Jensen Huang called Co-Packaged Optics essential for the AI build-out. What NVIDIA is doing is nothing short of extraordinary, not just for the future of AI, but for the American advanced manufacturing workforce, Corning CEO Wendell Weeks said in a press release. Corning's stock is up more than 250% in the past year as of Tuesday's close, driven by the 175-year-old company's rapid pivot into the new economy. In January, Meta announced it would spend up to $6 billion as the flagship customer, helping Corning build out its optical cable plant in Hickory, North Carolina, an expansion that's expected to create around 1,000 jobs. Analysts have long awaited NVIDIA's large-scale deployment of co-packaged optics because the technology promises to vastly increase the speed of data transfers and lower the energy needs for AI workloads. Corning is well known for making all the display glass for Apple's iPhone, but optical communications remains its largest and fastest-growing business. Since inventing optical fiber for long-range communications in 1970, Corning has provided millions of miles of cables to connect racks together in AI data centers from all the major players. By partnering with NVIDIA, Corning could be bringing glass fiber between the chips themselves eventually replacing the 5 copper cables inside the chip company rack systems like Verirubin Fiber optic cables are tiny bendable strands of glass that allow data to pass through as photons at far higher speeds with less energy than what used by traditional copper wires. Moving photons is between 5 and 20 times lower power usage than moving electrons, Weeks told CNBC in an interview in January. You're bringing the light conversion process right next to the computer chip, said Vlad Galibov, who covers enterprise infrastructure at research firm Omdia. Less power is wasted because now you're traveling a few millimeters, which requires far less energy than traveling across the circuit board. Galibov added that NVIDIA has pushed the entire ecosystem to innovate faster. Optical fiber also allows for less signal loss than copper, speeding up reliable communication and shortening the distance needed between the hundreds of thousands of GPUs in a data center. AI is driving the largest infrastructure build-out of our time and a once-in-a-generation opportunity to reinvigorate American manufacturing and supply chains, NVIDIA Zwang said in the press release. Together with Corning, we are inventing the future of computing with advanced optical technologies, building the foundation for AI infrastructure, where intelligence moves at the speed of light while advancing the proud tradition of made in America, end quote. Morgan Stanley has rolled out a crypto trading pilot on E-Trade, charging less than Coinbase, Robinhood, and Charles Schwab for your crypto trading ahead of a wider launch later in 2026. Quoting Bloomberg, Morgan Stanley's 50 basis point pricing is about half of Robinhood's, which starts at 95 basis points. Coinbase's fees start at 60 basis points, and Schwab said last month it will charge 75 basis points. Spot trading on E-Trade is just one of the inroads Morgan Stanley has made into the crypto space over the last year or so. It debuted a Bitcoin exchange-traded fund last month, the first Wall Street bank to do so, and it made it the cheapest fund in the category. It also has Ether and Solana ETFs on the way, and in February, it applied for a National Trust bank charter to allow Morgan Stanley to custody digital assets as well. Executives are readying an offering for clients to convert crypto into shares of exchange-traded products without having to sell the assets first, according to people familiar with the matter who asked not to be named discussing non-public information. On the institutional side, Morgan Stanley plans to add the ability to trade tokenized equities in the second half of this year. After the financial crisis, executives reshaped Morgan Stanley into a wealth management powerhouse less reliant on the traditional Wall Street business of trading and investment banking. Its $13 billion purchase of E-Trade in 2020 was a major push into the retail market, which came with a new suite of rivals in the digital brokerage space. One of them, Robinhood, was a financial technology startup able to move faster into crypto than federally regulated banks. It began offering crypto trading in 2018 and last year made $901 million in crypto transaction-based revenue, 20% of its annual net revenue. And that was just a fraction of Coinbase's haul. That firm pulled in $3.32 billion in consumer transaction revenue in 2025. Bitcoin and Ether transactions made up 45% of Coinbase's total trading volume last year. Coinbase survived industry upheaval a few years ago, including the collapse of rival FTX to become the biggest crypto exchange in the US. Now its roster of competitors has swelled to include traditional banks and other financial firms previously unable to do much in the space. It's going to be very competitive in the next couple of years, particularly given the regulatory moats are drying up, one analyst said, end quote. I mean, this would be one of those things that they are probably uniquely qualified to attempt. Google is apparently testing a personal agent codenamed Remy in the Gemini app that integrates with Google's services to take actions for users. So, an agent that works across your Gmail, your Google Drive, etc., etc., quoting Business Insider. Employees at the company have been testing the new AI agent internally named Remy, which runs in a staff-only version of Google's Gemini app and can integrate with a range of Google's other services, according to an internal document seen by Business Insider and two people familiar with the matter. Remy is your 24-7 personal agent for work, school, and daily life, powered by Gemini, reads a description of the agent. It elevates the Gemini app into a true assistant that can take actions on your behalf, not just answer questions or generate content Two people familiar with Remy said it is currently being tested by employees A Google spokesperson declined to comment Agents are a big focus for AI Labs right now As the underlying models have improved they are now better and more reliable at powering autonomous tools. Google doesn't yet have a widely available fully autonomous AI agent product. However, it has been rolling out agent mode and other related features that can perform multi-step tasks with access varying by subscription tier and region. Remy sounds more advanced than those other tools. In some ways, it sounds similar to OpenClaw, an AI agent that became a viral sensation earlier this year, and can perform tasks such as responding to messages or conducting research on behalf of users. In February, Sam Altman announced OpenAI was hiring OpenClaw's creator. An internal description of Remy states, deeply integrated across Google, Remy can monitor for things that matter to you, handle complex tasks proactively, and learn your preferences over time. It could not be learned whether there is a timeline for launching Remy to the public. However, the document describes Remy as a dogfooding project, which is a common practice at tech companies in which employees test products before launching them to users. The company will hold its I.O. event later this month, where it's expected to demonstrate its next wave of AI products. Agents will likely be a big focus then. Google DeepMind CEO Demis Hasibis has long talked about his vision to build a digital assistant. It could not immediately be learned why Google named the agent Remy. It has origins in the Latin word Remigius, meaning oarsman or rower, which would be fitting for an agent that's doing a lot of work, end quote. Sure, AI is everywhere, but that doesn't mean enterprise value is a given. In a recent survey, PwC found the amount of CEOs who reported revenue gains or cost reductions from AI is nearly equal to the amount who say they're still stuck. So what's causing the issues? PwC bullied it down to clarity. Leaders aren't clear about what's hype, what's reality, or where AI can actually create measurable impact. To help change that, PwC is offering their AI expertise and data. They explore how to tune out noise around AI and get clarity on what successful adoption looks like. Learn from the experts by heading to pwc.com slash US slash brew AI. That's pwc.com slash US slash brew AI. The information says that Meta is building an OpenClaw-inspired agent as well, internally called Hatch and powered by its MuseSpark model, and also an agentic shopping tool inside of Instagram. Quote, the aim is for Hatch, the name of which might change once it's launched, to be able to perform a range of tasks for people as part of its training work. For example, Meta has built sandbox web environments where the agent can be tested on simulations of real websites such as DoorDash, Etsy, Reddit, Yelp, and Outlook, the person said. Meta CEO Mark Zuckerberg has been intensely focused on the possibilities of AI agents seeing the technology as central to the vision for what he calls personal superintelligence. Meta's goal is to deliver agents that can understand your goals and then work day and night to help you achieve them, he said last week on the company's quarterly earnings call when the company also said it was increasing its capital spending on AI infrastructure this year to as much as $145 billion. He also described on that call some of the technical challenges to making an OpenClaw-like tool work for Meta's billions of users, requiring extensive infrastructure as well as smooth and easy use. OpenClaw caught fire this year with tech enthusiasts who use it to build their own autonomous AI agents, but it is far too complicated for most non-technical users. Meta tried to acquire OpenClaw earlier this year, its creator Peter Steinberg said on a podcast, but the agent tool was ultimately acquired by OpenAI in February. In its development so far, Hatch has been powered by Anthropics' Claude Opus 4.6 and Claude Sonnet 4.6 models. When launched, Hatch will be powered by Meta's latest AI model, Muse Spark, according to a third person familiar with the matter. Meta is ramping up efforts to improve Hatch's ability to decide when to take initiative rather than wait for instructions, the person familiar with Hatch said. The company is also expanding the amount of information the model can process at one time and improving its memory so it can retain details across conversations. At the same time Meta is refining how the agent responds to users and how it selects and uses tools Meta is racing against competitors that also are embracing agentic tools for consumers all driven by a similar belief that such tech will become the next major interface for how people shop, work, and interact online. The idea of shopping agents has been especially buzzy among big tech companies. Google in January launched Gemini Enterprise for customer experience, a platform that includes an AI-powered shopping agent capable of recommending products, building carts, and completing purchases with user approval. Amazon also offers an AI shopping assistant named Rufus, which helps customers track prices, research, and buy products. With the integration in Instagram, Meta hopes its agentic shopping tool can increase competition with TikTok's e-commerce feature, TikTok Shop, according to a fourth person familiar with the plans. Hatch users will be able to tap on an item in an Instagram Reel or Feed to learn more about it, navigate to an external page, and complete a purchase within the platform, the person said. The tool will build on new AI-powered shopping experiences that Meta announced in March, which included features that surface more detailed product information with AI and a new checkout experience that lets users make a purchase by just clicking an ad. Shopping agents face a number of hurdles, including resistance from some platforms to allowing other companies' agents to operate on their platforms. CEO Andy Jassy said last week that the experience with third-party agents on Amazon hasn't gotten great yet, noting they are not often able to get the pricing right or the product information right. Meta and other tech companies will need to persuade consumers that agents are dependable in the wake of incidents in which agents have made errors such as providing incorrect technical advice, end quote. finally today from the sounds boring but could end up being a big deal eventually file quoting the deep view open ai is getting creative to deal with the ai industry's imminent compute crunch on wednesday the chat gpt maker and a coalition of researchers from amd broadcom intel microsoft and nvidia published a paper offering a rare look into the company's training stack, debuting a new compute networking protocol designed to make GPU clusters faster, more reliable, and conserve precious compute cycles. The protocol, which has been in the works for two years, is instrumental in OpenAI scaling the compute that it needs to continue building bigger and better models, noting in a blog post that the networking approach accelerates its vision for Stargate, the company's long-term effort to garner the compute it needs to build and scale cutting-edge AI. The paper introduces a protocol called MRC, or Multipath Reliable Connection, which essentially tackles two main issues with the networks that serve the connective tissue of AI infrastructure, congestion and failures. As GPU clusters grow, these are problems that become more arduous to solve. This protocol relies on so-called packet spraying, said Mark Hanley, OpenAI's networking lead, which essentially scatters data along hundreds of paths in the network simultaneously to prevent any one network link from getting congested. This also reduces the amount of tiers in a GPU cluster, resulting in flatter networks that use up less of the data center's compute and power. When handling network failures, MRC detects and reroutes when paths go down in microseconds. This allows GPU clusters to continue training seamlessly even if parts of the network break down. Jumping in here to editorialize sounds like the internet. Additionally, the MRC pairs with a protocol called SRV6 or IPv6 segment routing, which essentially tells data the exact path it needs to take through a network rather than forcing the network switches to do the routing work themselves, further reducing the energy requirements of these switches and the data center more broadly. We want to use as much compute as we can get, but we also want to make sure that we're getting it efficiently and effectively, and this is a critical component of that. Greg Steinbrecher, OpenAI's workload lead, told the DeepView in an exclusive interview, the protocol is already in use in OpenAI and Microsoft's largest training clusters, including the Oracle site in Abelene, Texas, and in Microsoft's Fairweather supercomputers, and has been used to train multiple OpenAI models, end quote. Nothing more for you today. Talk to you tomorrow.
