Video Friday: Two Birotors Make a Quadrotor

We have spent the last decade obsessed with the software layer of the future. We built LLMs, we scaled the cloud, and we convinced ourselves that the hardest problems were already solved by silicon. But as I look at the recent breakthroughs coming out of labs like NYU Tandon and MIT, I am reminded of a fundamental truth: the real world is messy, heavy, and unforgiving. Code doesn't have to worry about gravity or friction, but the next generation of builders certainly does.

The Weight of Reality: Why Physics Trumps Hype

Lately, the conversation around robotics has been dominated by the "humanoid arms race." Companies like UBTECH are bragging about rolling a humanoid off the assembly line every ten minutes. That is an impressive feat of manufacturing, but for those of us actually building in this space, it raises a glaring question: what are they actually going to do? Producing ten thousand units of a machine that lacks a clear, high-utility purpose is just a very expensive way to fill a warehouse.

The real innovation isn't happening in the mimicry of human limbs; it is happening in the manipulation of physics. Take the NYU Tandon researchers and their "WorMa" robot. Instead of trying to build a complex, multi-jointed leg system that breaks every time it hits a 20-degree incline, they used something much simpler: water. By shifting liquid between its head and tail, the robot changes its center of gravity to gain traction or clear obstacles. It is an elegant solution to a mechanical problem. It doesn't look like a person, but it solves the problem of locomotion with 33% more efficiency than traditional methods.

The Founder Perspective: Simple Over Sophisticated

If you are a founder in the robotics or AI space, there is a lesson here. We often default to the most complex solution because it looks better in a pitch deck. A humanoid robot looks like the future; a plastic tube pumping water looks like a high school science project. But the plastic tube works. It handles inclines that would make a bipedal robot topple over. For builders, the goal should always be to solve the friction point with the least amount of mechanical complexity possible. The more moving parts you have, the more points of failure you have to maintain.

We see this same trend in the work being done at the Norwegian University of Science & Technology. They are taking quadrupeds—essentially robotic dogs—and making them amphibious. They aren't reinventing the wheel; they are machining waterproof housings from polyoxymethylene and using off-the-shelf O-rings. They are using standard SO(3) control theory to manage attitude underwater. This is what I call "commodity innovation." You take what exists, harden it for a new environment, and suddenly you have a tool for disaster response or environmental monitoring that actually works today, not in five years.

The Rise of Physical AI

We are seeing the term "Physical AI" pop up more frequently, and it’s more than just a buzzword. It represents the bridge between the digital brain and the mechanical body. The Kubi 2.0 project is a good example of this. It’s a simple desktop telepresence device, but it allows an AI agent to actually have a physical presence in a room. This is the low-hanging fruit of the industry. Before we have robots doing our laundry, we will have robots that simply allow us to be "present" in a location three thousand miles away.

However, we have to talk about safety. Agility Robotics is leaning into this with their Digit 5 model. It isn't a "pretty" robot, but it looks like a piece of industrial equipment. That distinction matters. If we want these machines in our homes or our hospitals, they need to look and act like safe tools, not experimental toys. The skepticism I hold toward the "humanoid in every home" narrative stems from this: we haven't yet proven that these machines are safe enough to be around children or the elderly without a handler nearby.

Ecosystem Engineering: Robots for Bees

Perhaps the most fascinating development isn't about robots helping humans, but robots helping the environment. The RoboRoyale project, discussed at CMU, is using micro-robotics to support honeybee colonies. This is a builder-first approach to a global crisis. Instead of trying to replace bees with drones—an incredibly difficult task—they are building systems to observe and support the existing biological infrastructure. They are tracking the Queen's behavior and mapping the health of the hive using AI to intervene only when necessary.

This is a shift from "replacement technology" to "augmentation technology." As a founder, you should ask yourself if your product is trying to replace a complex biological system or if it is providing the data and support to make that system more resilient. The latter is almost always a more sustainable business model.

The Takeaway for Builders

The hardware world is currently split into two camps. There are the manufacturers building thousands of humanoids, hoping the software catches up to make them useful. Then there are the engineers at places like MIT, building robotic optics labs that can autonomously run experiments to find new materials for solar cells. One is a bet on a vision; the other is a tool that accelerates the pace of discovery today.

My advice? Be the person building the tool. Whether it is a water-weighted worm that can climb stairs or a robotic arm that automates a lab, the value lies in solving specific, physical bottlenecks. The hype will always favor the humanoid, but the market will always favor the machine that gets the job done without breaking.

The goal isn't to build a machine that acts like a human; the goal is to build a machine that solves a problem humans are too inefficient, or too fragile, to handle.

We are entering a phase where the "cool factor" of robotics is wearing off, and the "utility factor" is taking over. For those of us looking at this from a founder's perspective, that is the best news we've had in years. It means the winners won't be the ones with the best marketing, but the ones with the best grasp of physics and the most reliable hardware.

Read the original at IEEE Robotics →