Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
0:00
|
8:19
The AI industry's most powerful executive was attacked twice in a single weekend, but the piece that may do the most lasting damage was published in a magazine. Across the Atlantic, two governments just declared emergencies over an AI model nobody outside Anthropic has been allowed to touch. And a country of 1.4 billion people just made the single biggest educational bet in AI history. Plus: the most prestigious AI conference of the year delivered a verdict on who's winning that nobody saw coming a year ago. It's a Monday that will be hard to forget.
Feedback? Email mike@yesterdayinai.news or connect on LinkedIn, X, Bluesky, or Substack. If you like the show, please take a minute to rate and review it so others can find it!
Hi folks, this is Yesterday in AI, your daily digest of everything happening in the world of artificial intelligence. I'm Mike Robinson. It's Tuesday, April 14th, and before the weekend could finish sorting through one crisis, the AI industry walked into Monday facing three more. Sam Altman's home was attacked a second time. The New Yorker's damaging profile of Altman from last week is still reverberating, and Britain's financial regulators declared an emergency over an AI model that hasn't even been released yet. Let's get into it. We covered the Molotov cocktail at Altman's home on Friday in yesterday's weekend catch up edition. Today the story deepened. Early Sunday morning, less than forty-eight hours later, a Honda Sedan drove past Altman's home and a passenger fired a shot. Two suspects were arrested, with three guns recovered at their residence. Nobody was hurt, but two attacks, two separate sets of suspects in the same weekend is not a coincidence. It's a pattern. Court documents released Monday add Critical Context. The first attacker had posted on Pause AI's Discord about fears of human extinction from AI. When FBI agents searched his home on Monday, they found a document describing opposition to AI, warnings about humanity's impending extinction, and a list of other AI technology executives. He wasn't angry at Altman specifically. He was angry at an industry, and Altman is its most visible face. Advocacy groups that warn about AI risks condemned the violence. What this weekend established is that the backlash against AI has crossed from online rhetoric into physical threat. The faces of this industry are operating in a different security environment than they were even a year ago. And then there was the New Yorker. The magazine published a profile of Altman a week ago that was 18 months in the making, drawn from more than 100 interviews. It is not flattering. Sources used words like liar and sociopath. The piece reports that Altman has confused basic machine learning terminology in front of his own engineers, that he dropped out of Stanford's computer science program, and that he has compared himself to J. Robert Oppenheimer, the physicist who helped build the atomic bomb and spent the rest of his life wrestling with what he'd created. The implicit question in all of this is about trust. OpenAI's technology is being threaded into healthcare, education, and government infrastructure, and the public now has a significant sourced reason to debate whether the person steering it is being straight about what he knows and what he's building. Now to the Mythos story, which also took a significant turn. Anthropic co-founder Jack Clark said publicly that the company is in active discussions with the Trump administration about Mythos, the unreleased frontier model too dangerous, by Anthropic's own description, to ship publicly. This matters because the Pentagon designated Anthropic as a supply chain risk last month, and a federal appeals court last week declined to block that designation while litigation proceeds. Anthropic is frozen out of the fastest growing government AI market as previously covered. Clark's message was pointed that contracting dispute should not obscure Anthropic's strong commitment to national security, and that government officials need to be informed about mythos and the models that follow. This is the new architecture of frontier AI governance, model by model, agency by agency negotiations about specific capabilities. Britain is reaching the same conclusion from a different direction. Following the Mythos announcement, the Bank of England, the Financial Conduct Authority, and His Majesty's Treasury are coordinating an emergency response with plans to consult the National Cybersecurity Center within two weeks. This echoes the Fed and Treasury briefing with bank CEOs we covered Friday. Regulators on both sides of the Atlantic are treating a single unreleased AI model as a present-tense risk to financial infrastructure. That kind of coordinated urgency over a product that isn't even publicly available is genuinely unprecedented. Let me shift to something more accessible. Anthropic launched Claude for Word today, a direct Microsoft Word integration where Claude lives inside your document, reads comments, makes tracked changes, drafts from templates with cited sources, and runs document-wide consistency checks. It also introduces reusable workflow skills for tasks like contract review that could be shared across teams. Claude for Word is currently in beta for team and enterprise subscribers. This follows Claude integrations with Excel and PowerPoint, completing the core office triumvirant, and it signals a clear product strategy. Rather than asking enterprise users to adopt a new interface, embed Claude into the tools where work already happens. The competitive implication is real. If you're a knowledge worker whose entire day runs inside Word, Excel, and PowerPoint, and Claude is already there anticipating your next move, the incentive to go open a separate chatbot window drops to nearly zero. Given that Anthropic posted a $30 billion annualized revenue run rate last week, the approach is clearly working. The coding agent race is also heating up on all fronts this week. Anthropic is testing UltraPlan, a cloud-based task planning system that moves planning out of the local terminal and into a collaborative browser interface with threaded comments and multi-repository support. Alongside it, Coordinator Mode would let Claude orchestrate high-level planning while delegating to parallel subagents. OpenAI is simultaneously testing Codec Scratchpad for parallel task execution, pointing toward a unified Codec SuperApp, and XAI is preparing a credits-based pricing model for Croc build with Model Arena, a multi-agent comparison feature. Three companies, three overlapping moves in the same week. The multi-agent parallel task coding environment is clearly where the major labs think the developer future lives. In other news, Apple is reportedly testing at least four smart glasses designed simultaneously, options ranging from rectangular Wayfarer style frames to slim circular variants in colors like blue and light brown. The defining decision? No display. The glasses will rely on cameras, microphones, speakers, and an AI-enhanced Siri for hands-free media capture, calls, and real-world processing. Think Meta's RayBan model, but built around Apple's privacy forward position and deep iPhone integration. Apple is targeting a reveal by late 2026 with consumer availability in 2027. The wearable AI market is shaping up to be the next platform battleground. Meta already has millions of RayBan units in the wild, and Apple entering that space with its hardware ecosystem and distribution would change the game entirely. One story I don't want to get lost in all the drama. China this week made AI education mandatory for every student in the country, from primary school through university. Five government ministries signed off together. AI joins the standard curriculum at every level, and it gets added to teacher certification exams. Smart systems will be integrated into classrooms for grading, tutoring, and lesson preparation. Full deployment by 2030, aligned with China's 15th five-year plan, which explicitly targets global leadership in AI applications. The U.S. debate right now is largely about what AI does to the workforce that exists today. China is engineering the workforce for the decade they want after that. Mandatory AI education at every level, backed by five ministries, is in a pilot program. It's national infrastructure investment on a 20-year timeline. Whatever your view on the geopolitics, the scale of the ambition deserves your attention. To close, for anyone reading the industry's tea leaves, the Human X conference in San Francisco last week offered a notable data point. Venture capitalist Roseanne Winseck said publicly, last year OpenAI looked like the clear winner. This year Anthropic feels miles ahead. Investors cautioned that momentum in this space can reverse quickly, and it has multiple times. But for a week that started with a New Yorker expose, two attacks on its CEO's home, and the most sustained negative press OpenAI has faced in years, the room temperature at the industry's biggest conference was not tilting in their direction. That's all for this edition of Yesterday in AI. Stay curious, and I'll see you tomorrow.
