The Body Keeps the Signal

A new preprint describes a soft pneumatic actuator printed in one step from conductive thermoplastic polyurethane. Its origami-like air chamber bends, while a connected sensing structure changes resistance under load. The authors report 40 degrees of bending, 12.5 newtons of output force, and a 26.9% resistance change across a 0–45 newton applied-force range. A two-finger gripper demonstrates the combined functions. The useful move is not simply adding a sensor: fabrication integrates sensing and motion without bonding dissimilar materials. These are laboratory results, not established durability or production readiness. Air supply and electrical measurement still sit outside the printed body.

Tactile-JEPA treats an electronic skin as a spatial arrangement rather than a camera image. Its self-supervised training hides groups of sensing elements and learns to predict their representations from the rest, using the connectivity of the surface. Local and wider masking help retain both contact detail and the state of the whole skin. Across three datasets, the authors report 6.3% lower force-estimation error and 20.8% lower in-hand orientation error than their prior-state-of-the-art comparison. The preprint tests magnetic and piezoresistive sensors across different configurations. This supports a learning method, not a universal skin that transfers unchanged to every body. Geometry becomes useful prior knowledge, but calibration and task-specific evaluation still matter.

Elastic metamaterials can guide vibration through carefully arranged geometry, but manufacturing variation does not disappear because a design is called topological. A new mathematical framework models continuous uncertainty in the dimensions and material properties of one-dimensional waveguides. The authors connect a time-varying formulation to established spectral-element methods, then extend it with stochastic fields and Monte Carlo analysis. Numerical studies cover rods, shafts and beams, including effects on topological bands. The practical contribution is a way to ask how much variation a proposed behavior can tolerate before making the object. It is a modeling result, not a manufactured product test or a guarantee of defect immunity.

A theoretical coding paper asks whether a printed object can retain its identity after it breaks. The proposed codes repeat multidimensional patterns through an object so an adequately sized fragment can identify the original codeword despite an unknown offset. The authors derive bounds and constructions, including versions that tolerate substitution errors. Crucially, “any fragment” has conditions: it must contain a box meeting specified volume and thickness requirements. This is not a demonstrated scanner or a guarantee that an arbitrary chip is readable. For physical-digital design, the interesting possibility is provenance distributed through matter instead of entrusted to one removable label. The privacy and ownership rules would still need deliberate design.

A preprint reconstructs the long-term development of U.S. road vehicles and separates installed capacity from the activity it actually supports. Its analysis distinguishes professional and consumer fleets, showing that power, mass and fuel use do not move together. A usage-factor measure helps describe engine oversizing and fleet redundancy: more capability can mean more material tied up without a corresponding increase in useful activity. The authors argue that avoiding oversized power capacity could reduce material demand during electrification. This is a system-level reconstruction, not a controlled comparison of individual vehicle models. It adds a necessary counterweight to the week’s integrated-function research: carrying more functions is valuable only when the object earns the resources it contains.