I was today days old when I learned there's a swimming robot out there smaller than a stick of gum that moves because part of it is, genuinely, alive. No batteries. No onboard electronics. Just a thin gel skeleton, a layer of real muscle cells on top, and someone shining a light on it from outside. MIT's Raman Lab built the thing, and the only way to drive it is to flash it.
Here's what it actually is, how you steer something with no controls at all, and why a lab at MIT is building robots out of living tissue on purpose.
What it actually is
The skeleton is a gel called GelMA (gelatin methacrylate), about half a millimeter thick, with square-bottomed grooves stamped into both sides and two fins extending off the body. The whole gel skeleton is about the size of a stick of gum. On top of it sits a layer of real muscle — genuine mouse muscle cells, grown directly onto the gel, and much thinner than a single strand of hair (reported elsewhere as under 15 micrometers). The cells have been genetically engineered using optogenetics, meaning they're built to twitch and contract in response to flashes of light rather than to a nerve signal or an electrical current.
How it's controlled — no battery, just light
There's no electronics package hiding inside this thing, and there's no battery to run down. Shine light on one fin and it beats; alternate the light between the two fins and the robot turns, the same way a real swimming animal steers by flexing one side more than the other. In testing, the researchers didn't even build a fixed light rig — they moved the light source by hand to steer the robot through the water, which is as close to "driving" a living machine as it gets right now.
The maze demo, and how fast it actually moves
At top speed, the robot can travel about four body lengths per minute, and the team demonstrated it navigating a watery maze inside a petri dish — small-scale, but a real test of whether light alone could actually steer it through turns, not just push it forward. Source: MIT News.
Who built it, and why use living tissue at all
The work comes out of MIT's Raman Lab in the Department of Mechanical Engineering, led by Ritu Raman, associate professor of mechanical engineering. The paper, titled "2D skeletal muscle thin film actuators enhance efficiency of biohybrid robots," published September 29, 2026 in the journal Advanced Functional Materials, with support from the Office of Naval Research. Maheera Bawa is the first author, alongside co-authors Arielle Berman, Laura Schwendeman, Ferdows Afghah, and Seanbiron Johnson. Raman's own reasoning for building with muscle instead of motors: living tissue is "soft, responsive to its surroundings, and can heal itself" — qualities a rigid robot doesn't get for free. Sources: ScienceDaily and SciTechDaily.
The Halloween-Season Honesty Check: This Isn't Reanimation
It's hard not to think of Mary Shelley's monster here — a lab bringing muscle to twitching life with a jolt of light instead of a bolt of lightning. But the honest version of the comparison breaks down exactly where it should: Frankenstein's creature was stitched together from dead tissue and shocked back to life. Raman Lab's muscle was never dead. It's living, genetically engineered mouse muscle, grown and cultured the normal way, then wired — biologically, not electrically — to respond to light instead of a nerve signal. No reanimation, no monster — just real living tissue that happens to take its stage directions from a flashlight instead of a brain. Spookier premise, tamer reality; the real science is the more interesting of the two anyway.
Why it matters — and why it's still early
The pitch isn't that this thing is ready to ship anywhere. It's a petri-dish maze demo, not a field deployment. But the stated potential is real: a robot built from self-healing, living muscle could do environmental monitoring in water, and it could explore fragile, unpredictable ecosystems that a rigid machine might otherwise damage just by being rigid. That's the whole case for building it out of muscle instead of motors in the first place — and right now, it's a promising early step, not a finished tool.