Packy McCormick's latest dispatch from Not Boring cuts through the usual noise of artificial intelligence hype by pivoting sharply from chatbots to the physical world. While much of the industry obsesses over text generation, McCormick argues that the real inflection point is happening in the realm of embodied cognition—where software finally learns to move, manipulate, and navigate the messy reality of our streets and homes. This is not a story about a single breakthrough, but a convergence of open-source robotics, drone logistics, and regulatory shifts that suggests the era of the "moving room" is closer than we think.
The Democratization of Physical Intelligence
McCormick opens with a striking observation about the shifting incentives in the tech sector: "the big tech companies came out in favor of open-weight models out of pure self-interest." He uses this to frame the emergence of Enigma, a stealth company that has surfaced with $71 million in funding and a bold experiment: allowing anyone to control over 100 real robot arms via a browser. The founders, Jonathan Jacobi and Gal Niv, are not traditional roboticists but former intelligence operatives from Israel's Unit 8200, a unit known for its cyber warfare capabilities. McCormick notes, "Neither is a roboticist. That might be a good thing - in deep tech, people can often get stuck in whichever path they focused on for their PhD."
This outsider perspective is central to the piece's optimism. By letting the public play with the robots—telling them to paint, defuse bombs, or run experiments—Enigma is gathering a unique dataset on how humans naturally interact with machines. McCormick writes, "If the race in robotics is about accumulating real training data fast and cost-efficiently, this is a novel and orthogonal approach to that which doubles as a unique marketing stunt." The argument here is that the bottleneck in robotics isn't just better hardware, but better interfaces and a deeper understanding of human intent. Critics might note that relying on public crowdsourcing for safety-critical training data carries inherent risks, but the sheer novelty of the approach forces the industry to rethink how models are trained.
No PhD, no stuck. Sometimes the best way to solve a deep tech problem is to ignore the established playbook.
From Tabletop to Whole-Body Control
The commentary then shifts to the massive strides made by Google DeepMind with its new Gemini Robotics 2 model family. McCormick highlights the move from simple tabletop manipulation to complex, whole-body movement. He describes the system as a "vision-language-action model that maps vision and language into motor control," capable of handling tasks like tying knots or sealing ziplock bags across different robot bodies. The significance lies in the unification of perception, planning, and control into a single model family.
"DeepMind says the same model checkpoint controlled multiple embodiments, including Apollo variants and a Franka Duo setup," McCormick points out, emphasizing the flexibility of the new architecture. This is a crucial step toward the concept of embodied cognition, where the robot's understanding of the world is inseparable from its ability to act within it. However, the author is careful not to overstate the current capabilities, noting that "movement speed still needs work, and the benchmarks show uneven performance on multi-finger dexterity." The progress is real, but the gap between a lab demonstration and a reliable household helper remains wide.
The Logistics of the Future Sky and Street
The piece takes a pragmatic turn as it examines the commercialization of autonomous delivery. DoorDash has secured FAA Part 135 air carrier certification, a rare credential that allows it to operate drones as an air carrier. McCormick interprets this not just as a delivery play, but as a strategic move to "commoditize its complements." By building the infrastructure for drone delivery, DoorDash is betting that it owns the most valuable part of the ecosystem—the network of customers and merchants—and can choose the most efficient delivery method regardless of the technology.
"The company's strategy and bet is clear here: they believe that it already owns the most important part of the delivery ecosystem... and that it can commoditize the ways that it gets things from one party to the other," McCormick writes. This logic mirrors the approach of other tech giants who build platforms to lower the barrier for competitors while securing their own position. Yet, the public reception remains skeptical. McCormick highlights a viral reply to DoorDash's announcement: "We all are supposed to just shut the fuck up and accept the retarded dystopia future they are creating for us." This tension between corporate efficiency and public anxiety about a drone-filled sky is a recurring theme that the author acknowledges but ultimately views as a hurdle to be cleared rather than a stop sign.
Similarly, the article celebrates Amazon's Zoox for receiving the first U.S. approval for a steering-wheel-free robotaxi. This vehicle is designed from the ground up without a driver's seat, featuring inward-facing seats that turn the car into a "moving room." McCormick finds this particularly compelling: "I can't wait to see the horrendous things people do to and in these cars." The implication is that the removal of the driver fundamentally changes the purpose of the vehicle, shifting it from a tool of transport to a space for living and working.
The Battle for the Narrative
Finally, McCormick addresses the broader strategic landscape of AI development, focusing on the rivalry between OpenAI and Anthropic. He frames OpenAI's recent decision to give 10,000 researchers free access to frontier models as a calculated counter-positioning move. "The best way to do this is for us to empower scientists, not to try to figure out everything ourselves," he quotes from Sam Altman's team. McCormick argues that this strategy is designed to "deflate fears that AI will replace us and to make Anthropic look weird."
The author suggests that market incentives are driving companies toward outcomes that benefit the public, even if the motives are competitive. "What's amazed me most this week, and made me appreciate capitalism, is that even if you don't trust Sam or OpenAI or big tech, companies are incentivized to do things that are good for the public, too," McCormick writes. This is a pragmatic, if slightly cynical, take on the industry: the fight for market share is inadvertently accelerating the democratization of powerful tools. A counterargument worth considering is that this "empowerment" is still tightly controlled by the companies that own the models, and true open access remains elusive.
Bottom Line
McCormick's piece succeeds by grounding the abstract hype of AI in the tangible reality of moving metal and flying drones. The strongest argument is that the next leap in intelligence will be physical, driven by the convergence of open-source data and whole-body models. The biggest vulnerability in the narrative is the assumption that public skepticism can be overcome simply by making the technology more efficient; the social contract for autonomous systems is far from settled. Readers should watch how regulators balance the rapid deployment of these technologies with the genuine fears of a public facing a rapidly changing physical world.