MIT engineers built a swimmer whose engine is a layer of living cells grown on a gel film the thickness of a few sheets of paper. It is slow, tiny, and steered by a flashlight. That is exactly why it matters.
Every robot you have seen is an argument about energy. Where does it store it, how does it convert it, how much does the converter weigh? Motors, gearboxes, batteries and wires are the answer we inherited from the twentieth century, and we have spent decades making them smaller.
This week's cover story proposes a different move. A team at MIT coated a gel film with muscle cells engineered to contract when light hits them. No motor. No onboard electronics. The robot swims when you shine a light on one fin. Its top speed is about four body lengths per minute, which is leisurely. If you stop there, you miss it.
Here is the lens for the issue: ask what a technology no longer needs, not how fast it goes. A motor needs a factory. This actuator needs a nutrient bath and, as the researchers showed, it gets stronger when you exercise it. Elsewhere in these pages: five silicon qubits that stay accurate together, a painted-on treatment that halts childhood decay, and a rover that out-drove its predecessor in a third of the time. Read each for what it dropped.
— The Editors
A half-millimetre film of engineered muscle just did what bulky tissue chunks could not: push a robot through water.
Imagine you had to build an engine out of something that is alive. Not a metaphor: cells, in a bath of nutrients, that contract when told to. That is the bet of biohybrid robotics, and for years it has had a sizing problem. Muscle is a wonderful actuator in a body, but in a lab it was grown as thick 3D lumps that needed millions of cells and were hard to keep alive. This week, ScienceDaily and others reported MIT's answer: stop growing lumps and grow a film.
The team, led by Ritu Raman, associate professor of mechanical engineering at MIT, built a robot about the size of a stick of gum. It has two fins; each fin measures 15 mm long, 7 mm wide and 0.5 mm thick, according to New Atlas's report. On each fin sits a single layer of skeletal muscle cells, genetically modified so that light makes them contract.
Shine a light on the left fin and the left fin flaps. Light the right and the right flaps. Alternate them and the robot swims and steers. A researcher guided it through an underwater maze in a petri dish with a handheld light source. There is no battery to carry and no electronics on board.
Water is unforgiving. As Raman put it, "It takes a lot of force to move through water versus air." The team's earlier project, an artificial iris-like disk, produced cell movement of only about 100 microns, far too little to propel anything. Their gel base also shrank under the forces the cells generated, wasting the work.
The fix was less glamorous than a new cell type. It was materials engineering: swap the soft fibrin gel for a stiffer gelatin methacrylate, and stamp grooves into it so the cells line up like rowers in a boat.
Researchers describe it as the first example of a very thin, two-dimensional, muscle-powered robot capable of locomotion.
The team moved from fibrin to gelatin methacrylate (GelMA). Stiffer gels produced better cell alignment and stronger contractions, and they stopped shrinking and wrinkling as the muscle pulled against them. The base film is about half a millimetre thick.
Square-bottomed grooves aligned the cells better than curved-bottom channels. Better alignment helped the cells fuse into coordinated fibers that pull in the same direction, which is the difference between a crowd and a rowing crew.
Before the robot was assembled, the muscle was trained with repeated pulses of light. According to New Atlas, that meant 15 minutes of light stimulation a day at two pulses per second, and it increased fin displacement fourfold compared with an unstimulated control group.
Think about that for a moment. Nobody trains a motor. A motor wears out with use; this actuator improved with it.
The muscle kept working for more than 30 days without being anchored, per New Atlas. The competition is not standing still: another 2026 design, OstraBot, reached 15.6 body lengths per minute, about four times the MIT robot's top speed. The MIT team's own stated goal is simply to prove that a very thin engineered muscle layer can generate enough force to swim.
The authors note that earlier biohybrid robots used bulky 3D muscle chunks requiring millions of cells, while the thinner design may need fewer. That is a hypothesis, not yet a result.
Let us be disciplined. This robot is not ready for the open ocean. It is slow and tiny, it was tested in a lab dish, and someone outside the dish has to aim the light. The paper's own framing is a demonstration of force, not a product.
The researchers point to environmental monitoring in water and work in fragile settings where rigid machines could damage the surroundings. Raman notes that living tissue is soft, responsive to its surroundings, and can heal itself. Those are the properties a motor cannot offer, and they define where this class of device could eventually compete.
The strategic takeaway for a technology leader: the engineering effort here went into interfaces, namely the gel, the groove geometry and the training regime, rather than into a better engine. When the component is alive, the system design is the product.
Funding came in part from the Office of Naval Research, which tells you who is watching for underwater applications. The next goal, per the team, is optimizing body design for faster swimming.
Biohybrid robot — a machine combining living tissue with synthetic structure.
Actuator — the part that turns energy into motion; here, muscle.
GelMA — gelatin methacrylate, a tunable gel used as the robot's skeleton.
Body length per minute — speed normalized to size; MIT's robot manages about 4.
Everyone who sees this robot will grade it on speed, and it fails: four body lengths a minute, beaten about four to one by OstraBot. The contrarian read is that speed is the wrong axis, because the interesting claim is about the bill of materials.
No motor. No battery. No onboard electronics. The control signal is photons from outside; the engine is grown from cells, improves with 15 minutes of daily exercise, and kept working for 30-plus days unanchored. A roboticist's instinct is to optimise the engine. This team optimised the substrate and let biology supply the engine.
Our hypothesis: the commercial wedge for biohybrids is not "slower robots" but places where a motor is a liability, such as tiny scales and fragile environments. In those places, a competitor with a better motor is not a competitor at all.
The skeptics are not wrong. The robot is gum-stick sized and was tested in a lab dish, not natural environments. It is steered by a person with a light. Long-term durability beyond about a month was not reported. It uses one cell line, one fibre direction, simple shapes.
Living actuators also bring problems motors never had: they need a nutrient environment, they are difficult to standardise, and scaling cell culture is a manufacturing discipline of its own. The authors say plainly that their next step is faster swimming, which concedes that speed matters.
So treat this as an option, not a forecast. If a biohybrid ever beats a motor, it will be on a metric nobody puts on the spec sheet today.
RIKEN researchers operated five silicon spin qubits at once on a shared microwave line at about 99.9% single-qubit fidelity (above 99.99% with up to three active). Silicon qubits can be made with chip-industry methods, and the team found and compensated a key error source, the AC Stark shift. Honest caveat: it is still five qubits, short of fault-tolerant levels.
In a Phase III trial published in JAMA Pediatrics, a 38% silver diamine fluoride solution, painted on with a sponge-tipped applicator, stopped decay in more than half of affected baby teeth in 830 children under six. No drilling or sedation, though treated areas stay dark and the cavity use is still off-label in the US.
On October 11 NASA's Perseverance passed Opportunity's 28.06-mile mark to become the farthest-travelling vehicle on another world, in about five and a half years versus Opportunity's nearly 15. Its AutoNav system evaluates over 90% of its driving distance, versus about 5% for early autonomy on Spirit and Opportunity.
Look back across the issue. A swimmer without a motor. Qubits that share a control line instead of each demanding its own. A cavity treatment with no drill. A rover that increasingly decides its own route. Each is progress by subtraction.
Subtraction has a way of looking like weakness at first. Four body lengths a minute is a poor number until you notice the robot carries no battery. Five qubits is a toy until you notice they are made like chips. The discipline for a leader is to keep two ledgers: one for what a technology can do today, and one for what it has stopped requiring.
The second ledger is where options hide. Whether biohybrids ever leave the dish is unknown; the MIT team itself says the next step is simply swimming faster. But the question to carry into your own roadmap is portable: what is the most expensive thing you assume you must build, and who has just shown a way to grow it, share it, or skip it?