A small cluster of springs and metal bars, bolted together on a benchtop in Northfield, Minnesota, can count, distinguish odd from even, and remember how hard it has been pushed. It does all of this without a battery, a circuit board, or a single watt of electricity. The device, built by physicists at St. Olaf College and described in a paper published in Nature Communications in April 2026, is a working mechanical computer powered entirely by the snap of its own components.
“We’re exploring how everyday materials can hold and process information on their own,” physicist Joey Paulsen, who led the research, said in a statement released by the college. The project started from a deceptively simple question: how much computation can you squeeze out of physics alone, with no electronics involved?
How springs become logic gates
The core building block is what physicists call a “hysteron,” a spring-and-bar unit that snaps between two stable positions when pushed past a threshold force. Think of a light switch that stays up or down on its own. Each hysteron holds one of two states, functioning like a single bit in a conventional computer.
By linking multiple hysterons with geometry-tunable springs, Paulsen’s team created a network where flipping one element influences its neighbors. Adjust the spring geometry one way, and neighboring units tend to align in the same state. Adjust it another way, and they favor opposite states. Mix both types of coupling in a single device, and you can program how information flows through the structure.
The paper details a critical technical achievement: engineering non-reciprocal interactions between hysterons. In most mechanical systems, pushing element A affects element B the same way pushing B affects A. The St. Olaf team broke that symmetry, creating one-directional influence chains purely through spring geometry. That asymmetry is what allows the device to perform sequential logic, not just store a static pattern.
Three tasks, zero electricity
The team demonstrated three specific capabilities. First, the device counts: successive pulls cause a predictable sequence of flips across the array, and the resulting pattern of states encodes how many cycles have occurred. Second, it discriminates between odd and even inputs, landing in one configuration after an odd number of pulls and a different one after an even number, effectively running a parity check. Third, it retains what the researchers call “force-level memory,” recording not just whether it was pulled but how hard.
All three tasks emerge from the physical arrangement of the hysterons and their connecting springs. No software is involved. No stored electrical charge plays any role. A single external pull propagates through the network in a controlled way, and the device’s final configuration is the output.
What the research does not yet answer
The paper establishes that the concept works in a laboratory for a narrow set of tasks. Several important questions remain open.
Speed and reliability are not benchmarked. How fast the device can cycle through computations, and how consistently it performs over thousands of repetitions, will determine whether the concept moves beyond a physics demonstration. Durability is a related concern: the hysterons rely on repeated snapping, which subjects springs and joints to mechanical fatigue. The paper does not specify a cycle-life limit.
Scalability is the biggest unknown. The demonstrated tasks are elementary. Whether the platform can handle more complex logic, or whether friction, fatigue, and manufacturing tolerances impose hard limits, is not addressed. The paper focuses on establishing the fundamental mechanism rather than projecting practical applications.
No independent expert commentary has accompanied the published findings so far. Peer review at Nature Communications provides one layer of external validation, but broader reaction from the physics and metamaterials communities has yet to surface publicly.
Where a powerless computer could matter
If the concept scales, the most immediate applications would likely involve environments where electrical power is unavailable or impractical. A purely mechanical counter that logs how many times a threshold force has been exceeded, whether from wind gusts, wave impacts, or structural loads, could serve as a low-maintenance diagnostic tool in remote locations. No battery changes, no wiring, no solar panels required.
More broadly, the work fits into a growing field sometimes called “material intelligence,” where the properties and arrangement of matter perform information processing that would otherwise require electronics. Researchers in metamaterials and soft robotics have been exploring similar ideas, but the St. Olaf device stands out for demonstrating tunable, non-reciprocal mechanical interactions that enable sequential logic.
Computation without a plug
Nobody is suggesting that bistable springs will replace microchips. But the St. Olaf prototype makes a point that is easy to overlook in an era of trillion-transistor processors: computation is not inherently electrical. Wherever physics allows stable states and controlled transitions between them, information can be stored and processed. Paulsen’s team has shown that a handful of springs and metal bars, carefully arranged, is enough to make that happen.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
