Tech Brew Ride Home - Mon. 03/10 – Is Manus “DeepSeek” 2.0?

Episode Date: March 10, 2025

Well, do we have another DeepSeek moment on our hands? I tell you about Manus, which had some people losing their minds over the weekend. Now do we have the first signs of the AI M&A I’ve been looking for? Absolutely everybody wants to get in on the Stablecoin business. And the AI crisis at Apple delayed a product we might have seen this month. Sponsors: Kinsta.com/brian Timeline.com/ride Links: China’s Autonomous Agent, Manus, Changes Everything (Forbes) Manus probably isn’t China’s second ‘DeepSeek moment’ (TechCrunch) Undocumented commands found in Bluetooth chip used by a billion devices (BleepingComputer) ServiceNow Nears Deal to Buy AI Assistant Maker Moveworks (Bloomberg) Banks and fintechs join ‘stablecoin gold rush’ (Financial Times) Apple Clears the Decks With Rare Early-Year Product Frenzy (Bloomberg) Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 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 Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the Tech meme right home from Monday, March 10th, 2025. I'm Brian McCullough today. Well, do we have another deep-seek moment on our hands? I tell you about Manus, which had some people losing their minds over the weekend. Now, do we have the first signs of the AI M&A I've been looking for? Absolutely everybody wants to get in on the stable coin business, and the AI crisis at Apple delayed a product we might have seen this month. Here's what you missed today in the world of tech. Well, maybe you heard about Manus over the weekend. Some folks are calling this China's new Deep Seek moment. If so, the breakthrough here is that the creators claim this is the world's first truly autonomous

AI agent. Quoting Forbes, Manus represents something entirely different. It is not just another model. It is an agent, an AI system that thinks, plans, and executes tasks independently, capable of navigating the real world as seamlessly as a human intern with an unlimited attention span. Within hours, its March 6 launch would send shockwaves through the global AI community, re-igniting a debate that had simmered for decades. What happens when artificial intelligence

stops asking for permission and starts making its own decisions? From analyzing financial transactions to screening job candidates, Manus navigates the digital world without oversight making decisions with a speed and precision that even the most seasoned professionals struggle to match. In essence, it is a digital polymath trained to manage tasks across industries without the inefficiencies of human hesitation. This is what sets Manus apart from its Western counterpart, While ChatGTP4 and Google's Gemini rely on human prompts to guide them, Manus doesn't wait for instructions. Instead, it is designed to initiate tasks on its own, assess new information, and dynamically adjust its approach. For instance, given a zip file of resumes, Manus doesn't just rank candidates.

It reads through each one, extracts relevant skills, cross-references them with job market trends, and presents a fully optimized hiring decision complete with an Excel spreadsheet it generated on its own. When given a vague command like Find Me an apartment in San Francisco, it goes beyond listing search results, it considers crime statistics, rental trends, even weather patterns, and delivers a short list of properties tailored to the user's unstated preferences. Consider Rowan Chung, a tech writer who tested Manus by asking it to write a biography of himself and build a personal website. Within minutes, the agent had scraped social media, extracted professional highlights, generated

a neatly formatted biography, coded a functional website, and deployed it on a functional website, and deployed it online. It even troubleshot hosting issues without ever asking for additional input. For AI developers, this is the Holy Grail, a system that doesn't just generate information but applies it, fixes its mistakes, and refines its output. For professionals who rely on tasks, Manus can perform, it is an existential threat. But how did China, often perceived as trailing the U.S. in foundational AI research, produce something that Silicon Valley had only theorized about? And more importantly, what does it mean for the balance of power in artificial intelligence, end quote. Well, that is what people have been debating all weekend, because some early Manus users say the

agentic AI is no panacea with long waits, errors, unsatisfying answers, and endless loops often plaguing the experience. Quoting Dean W. Ball on Twitter, if an American firm had shipped Manus last month at Manus's current quality level, they'd currently be facing multiple investigations by state attorneys general, and if a Democrat had won the White House, FTC, and or DOJ2, and conceivably DOL, SEC, EEOC, pick your poison, end quote. Quoting TechCrunch, the head of product at Hugging Face called Manus the most impressive AI tool I've ever tried. AI policy researcher Dean Ball described Manus as the most sophisticated computer using AI. The official Discord server for Manus grew to over 138,000 members in just a few

days and invite codes for Manis are reportedly selling for thousands of dollars on Chinese reseller app, Gian Yu. But it's not clear the hype is justified. Alexander Doria, the co-founder of AI startup Play Us, said in a post on X that he encountered error messages and endless loops while testing Manus. Other X users pointed out that Manus makes mistakes on factual questions and doesn't consistently cite its work and often misses information that's easily found online. My own experience with Manus hasn't been incredibly positive. I asked the platform to handle what seemed to me like a pretty straightforward request, order a fried chicken sandwich from a top-rated food joint in my delivery range. After about 10 minutes, Manus crashed. On the second attempt, it found a

menu item that met my criteria, but Manus couldn't complete the order process or provide a checkout link, even. Manus wasn't developed entirely from scratch. According to reports on social media, the platform uses a combination of existing and fine-tuned AI models, including Anthropics Claude and Alibaba's Quen, to perform tasks such as drafting research reports and analyzing financial filings. A spokesperson for Manus sent TechCrunch the following statement via DM. As a small team, our focus is to keep improving Manus and make AI agents that actually help users solve problems. The primary goal of the current closed beta is to stress test various parts of the system and identify issues. We deeply appreciate the valuable insights shared by

everyone, end quote. So if Manus is falling short of its technical promises, why did it blow up? A few factors contributed such as the exclusivity created by a scarcity of invites. Chinese media was quick to tout Manus as an AI breakthrough. Publication QQ News called it the pride of domestic products. Meanwhile, AI influencers on social media spread misinformation about Manus's capabilities. A widely shared video showed a desktop program ostensibly Manus taking action across multiple smartphone apps. G confirmed that the video wasn't, in fact, a demo of Manus. Other influential AI accounts on X sought to draw comparisons between Manus and Chinese AI company DeepSeek comparisons not necessarily rooted in fact. The company behind Manus,

the butterfly effect, didn't develop any in-house models unlike DeepSeek. And while Deepseek made many of its technologies openly available, Monica hasn't, at least not yet. To be fair to the butterfly effect, Manus is in very early access. The company claims it's working to scale computing capacity and fix issues as they're reported. But as the platform currently exists, Manus appears to be a case of hype running ahead of technological innovation, end quote. This is quietly a huge deal. Tar Logic researchers have found an alleged undocumented backdoor in Chinese manufacturer Espresif's ESP 32 microchip used in more than one billion devices for Wi-Fi and Bluetooth connectivity. Quoting bleeping computer. The undocumented

commands allow spoofing of trusted devices, unauthorized data access pivoting to other devices on the network, and potentially establishing long-term persistence. The researchers warned that ESP32 is one of the world's most widely used chips for Wi-Fi and Bluetooth connectivity in Internet of Things devices, so the risk is significant. In total, they found 29 undocumented commands collectively characterized as backdoors that could be used for memory manipulation, read, write, RAM, and flash, Mac address spoofing, device impersonation, and LMP-LCP packet injection. Espresive has not publicly documented these commands, so either they weren't meant to be accessible, or they were left in by mistake.

The issue is now tracked under CVE 2025-27840, end quote. Now, the problem here is that the vulnerability isn't just in standard exploitation paths. It exposes systemic weaknesses at the OEM and supply chain. chain level where trusted relationships could be weaponized. While remote exploitation via Bluetooth HCI commands theoretically exist, the real industry concern is how inconsistently Bluetooth stacks handle these commands across the ecosystem. This creates an asymmetric risk profile across products using similar components. The mention of post-compromised escalation suggests a new attack vector where initial access can be leveraged into persistent hardware-level control, potentially bypassing

security updates. The final point subtly acknowledges that for targeted attacks against high-value tech executives or workers, physical access vectors remain more reliable than remote exploits. Something, something for a healthy startup ecosystem, we need to start seeing some M&A activity, as we've discussed previously. Workflow Automation Company ServiceNow has announced an agreement to buy Enterprise AI Assistant Company Moveworks in a deal that could value Moveworks at a around $3 billion. Quoting Bloomberg, the acquisition, Service Now's largest yet, is expected to close in the second half of 2025. Service Now said in a statement, created in 2016 MoveWorks provides

companies with AI assistance to deal with employee requests. Its technology is used by companies including Unilever, GitHub, and Broadcom, according to its website. MoveWorks has received backing from investors, including Kleiner Perkins, Lightspeed Venture Partners, and Bain Capital Ventures. The company was valued at $2.1 billion in a 2021 funding round. The majority of MoVWorks customer deployments already used ServiceNow, the company said. Initial integration between the companies will focus on offering an AI-powered self-service tool for customers, led by Chief Executive Officer Bill McDermott. ServiceNow makes applications that help firms organize and automate their personnel and information technology operations.

Like many of its peers, the Santa Clara, California-based company, has been baking generative AI features into its products. The acquisition is a sign of ServiceNow's potential deal-making ambitions under McDermen. Dermit. Prior to joining the company, he was CEO of Software Giant SAP, where he oversaw a number of high-profile deals. These included the multi-billion dollar purchases of Qualtricks, International, and Concord Technologies. McDermott previously signaled he'd focus on organic growth at ServiceNow. Service Now said in a statement that as of the end of 2024, it passed $200 million in annual contract value for its higher-priced product tier that includes AI features. It has nearly 1,000 AI customers, it said, end quote. everybody wants a piece of the stable coin action. Why? Because it's almost literally a license to print money.

According to the FT, Stripe, Revolut, and other banks and fintechs are all rushing to launch stable coins. There are around $210 billion in stable coins issued globally now, including Tether at $142 billion. Quote, last month Bank of America signaled it was open to issuing its own coin, joining established payments providers such as standard chartered, PayPal, Revolut, and Stripe in targeting a business dominated by cryptocurrency groups Tether and Circle. Their enthusiasm has been fueled by growing acceptance among regulators around the world that Stablecoins, designed to hold a constant value of a dollar per coin, could become a more accepted part of the financial system. That shift after regulatory hostility to META's Libra Stablecoin six years ago has been given further impetus by U.S. President Donald Trump's fervent embrace of cryptocurrencies. It's about people selling shoveling

in the stablecoin gold rush, said Simon Taylor, co-founder of FinTech Consultancy 11FS, who likened it to FOMO or fear of missing out. The other thing that's driven it is there's real volume, he said. Founders want to get a piece of it because they know they're going to get stablecoin regulation, and so it's all of those things coming together. Although stablecoins have typically been used to shift money between differing cryptocurrencies, they are growing in popularity in emerging markets as an alternative to local banks for payments, particularly in commodities, agriculture, and shipping. They are a type of private digital cash that acts as a de facto reserve of a sovereign currency, overwhelmingly U.S. dollars, and payments in digital coins

allow companies and consumers to cheaply and instantly access hard currency outside the banking system. There are about $210 billion worth of stable coins issued globally with about $142 billion printed by El Salvador-based Tether, and $57 billion by the U.S.'s circle, branded as USDT and USDC, respectively. Elon Musk's SpaceX uses them to repatriate funds from selling Starlink satellites in Argentina and Nigeria, while Scale AI offers its large workforce of overseas contractors the option of being paid in digital tokens. Transaction volumes climbed to $710 billion last month, compared with $521 billion in the same period a year ago, while the number of unique stable coin addresses has risen to $35 million, up 50% over the same period, according to data from Visa.

Large banks are growing increasingly confident pushing into the industry as regulations emerge. U.S. politicians are debating bills in Congress that set out standards for stable coins, giving banks, companies, and ordinary consumers more confidence to use the tokens. If they make that legal, we will go into that business, commented Brian Moynihan, chief executive of Bank of America on the Trump administration plans at the Economic Club of Washington last month, end quote. Saved Mark German Apple scoop Monday for last again today. his sources say Apple planned to announce a smart home hub this month, but postponed the launch since the device relies on the delayed Siri capabilities. So the AI crisis continues, quoting Mark.

After last week's product refreshes, Apple is now left with a clean slate. There aren't many device updates in the pipeline, perhaps excluding a new air tag, until later in the year when the iPhone 17 series, new Apple watches, and additional Macs are expected to arrive. But that leaves the question, and why was the company so eager to clear the decks? The answer is probably its worldwide developers conference in June. That event is always eventful for Apple, but this one promises to be one of the more consequential ones. A less cluttered schedule will help the company prepare for the event.

Apple is at a crossroads right now, and WWDC is a chance to convince developers, fans, and investors that the company is heading in the right direction, especially when it comes to AI. The iPhone maker just delayed Siri features that it promised at the previous WWDC renewing concerns that its artificial intelligence strategy is in disarray. The bad news is that Apple is unlikely to unveil groundbreaking new AI features at this coming WWDC. Instead, it will likely lay out plans for

bringing current capabilities to more apps. Apple is also still working on a foray into a new product category, the Smart Home Hub. As I've reported before, it's developing a device code named J490 with an iPad-like screen and home control features. At one point, the company had hoped to announce this product in March. But because the device, to an extent, relies on the delayed Siri capabilities, it has been postponed as well. In the meantime, Apple has opened up an internal testing program for the Home Hub, allowing select employees to take it home and provide frequent feedback. Meanwhile, Apple continues to explore ideas for smart glasses and other wearable devices. The Vision Pro headset, despite being a technical marvel, is clearly a commercial flop. Apple realizes that

it's not worth continuing to invest in this approach unless it's just to increase the devices specifications or reduce the cost and weight. So attention has shifted to figuring out a more appealing form factor. One intriguing idea was to build augmented reality glasses that serve as an external Mac display, but that was axed earlier this year. Instead, Apple is back to trying to get standalone AR glasses to work. We're talking about lightweight AI-infused spectacles that you could wear all day. It'll probably take Apple another three to five years to crack this challenge at best, but the company is exploring a lot of ideas and underlying technologies right now. Apple also is still discussing the idea of making its own version of the popular Rayband glasses from meta platforms.

Such a product wouldn't be a proper AR device, but it would include AI microphones and cameras to create a pretty good user experience. If Apple can bring its design prowess, offer AirPods-level audio quality and tightly integrate the glasses with the iPhone, I think the company would have a smash hit. It's mind-boggling that Apple hasn't gotten there yet. The Rayband glasses are clearly doing great in the marketplace and provide a stepping stone to AR. Apple also is still actively developing a product that would combine AirPods with cameras. The cameras would help power AI features by gathering information on the surrounding environment, as I've reported before. Of course, this all hinges on Apple getting its AI in order, end quote. Once again, my wife's professional connections came through this weekend.

We went to the first official dress rehearsal in front of an audience of the new Glenn Gell. Glen Ross Revival, which featured a Better Call Saul reunion of Bob Odenkirk and Bill Burr, who was excellent, by the way, I predict a Tony for him. But also, in the role of Richard Roma, the newly Academy awarded Kieran Culkin. It was really a great show, and as I think I've mentioned before, I've seen the movie Glenn Gary, Glenn Ross maybe over a hundred times in my life. It's my favorite movie by far, so I was sitting there mouthing along to the words, like I was at a concert for my favorite band or something. Well done, everyone. recommend going to see it if you can talk to you tomorrow