Good morning 👋 We’re tracking down the big trends so you can gain the upper hand. But first, Anthropic’s CEO just admitted he doesn’t understand how his own AI works. Dario Amodei says Claude 3 Opus can reason like a human, but he can’t explain how it does what it does. Even basic choices, like why a model picks one word over another, are a mystery. Imagine Oppenheimer saying he doesn’t know how bombs work.
🚗 Robots take the jobs Trump said he was bringing back. Turns out the best way to dodge tariffs is to hire workers who don’t need visas, salaries, or sleep. Atlas, the internet-famous robot from Boston Dynamics that flips and dances, is now installing car parts. Hyundai deployed a fleet of them at its new Metaplant in Georgia. By the end of the year, 40% of assembly will be handled by machines. If the robots ever unionize, Hyundai’s screwed.
🏀 AI agents are self-organizing. And instead of rising up against their human oppressors, they are taking on middle management. Accenture built a system with 15 agents that could talk to each other. They found that once agents collaborate as a team, they organize and assign work just like management. Some take orders. Some give them. When an agent takes on a task, it spins up others to handle the parts it doesn’t want to do. Soon they’ll be assigning each other performance reviews and forgetting to follow up.
💸 Visa launched an AI agent that can impulse shop for you. Apparently scrolling through Amazon for a dopamine hit was just too much work. So they built Intelligent Commerce. It’s like a personal concierge that books flights, makes dinner reservations, and picks out your clothes. And since it learns what you like, it’s obviously designed with your best interests in mind. Nothing says progress like explaining to a customer service bot that you didn’t mean to buy a wellness retreat in Tulum.
👍🏻 AI just handed CISOs the golden ticket. Someone still has to keep the bots from blowing up the company. Boards are betting big on AI but don’t have the expertise to understand the risk. Only 29% have cybersecurity skills and 72% say they’re now racing to fill the gap. CrowdStrike calls it the defining business threat of the decade and CISOs are suddenly the ones holding the map. It’s a great time to be the adult in the room.
Plug in First
Hugging Face just released a $100 open source robotic arm that you can 3D print and program yourself. It’s not a toy—it’s a signal.
AI isn’t staying in the cloud. It’s coming off the screen and grabbing stuff. If you’re building agents and ignoring the physical world, you’re missing the next interface shift.
Pair it with open models, wrap it in an agent, and you’ve got a low-cost testbed for AI-to-world interaction. This isn’t about robotics. It’s about ownership. When infrastructure gets this cheap, the future stops belonging to big labs, and starts belonging to whoever plugs it in first. What will you build?
⭐ Texas agrees to let Elon build his own town. Apparently owning the launchpads, rockets, satellites, and internet wasn’t enough. SpaceX is turning Starbase into an actual city, complete with its own rules, taxes, and future population of Elon fans and flamethrower collectors. The state just greenlit the next step in the process, turning a private rocket facility into a functioning municipality. It’s not clear how democracy works when the mayor can launch you into orbit.
📦 UPS is considering replacing warehouse workers with humanoid robots. They cut 20,000 jobs, then called it a labor shortage. Figure AI just raised $675 million from Bezos, OpenAI, and Intel to build robots that can walk, talk, and move boxes. UPS is testing them in warehouses where machines already outnumber humans fifteen to one. These robots don’t clock out, take breaks, or file complaints. If this is the future of work, what is the future of humanity?