Engineering: AI Engines & Solar Gains

A rocket engine is a controlled metal explosion. Fuel and oxygen burn in a chamber, the hot gas is squeezed through a nozzle, and the nozzle turns pressure into thrust that can lift tons of hardware off the ground. (sheffield.ac.uk) That sounds simple until you build one. A liquid engine has to survive extreme heat, route propellants through tiny channels, and keep combustion stable, which is why a new engine design usually takes months or years to model, machine, and test. (sheffield.ac.uk) LEAP 71 is trying to compress that cycle with software it calls Noyron, a “Large Computational Engineering Model.” In June 2024, the company said Noyron designed a liquid oxygen and kerosene engine autonomously, the part was 3D-printed in copper, and the engine was then hot-fired successfully. (leap71.com) The speed claim is the part engineers notice. The University of Sheffield, which worked on the test campaign, said the team went from preliminary design to a fully manufactured engine in “a matter of weeks,” with the design itself taking less than 2 weeks and print-to-hot-fire also coming in under 2 weeks. (sheffield.ac.uk) LEAP 71 pushed that idea further in December 2025. The company said it went from specification to first flame in under three weeks for two 20 kilonewton methane-and-liquid-oxygen engines, one with a conventional bell nozzle and one with a full-scale aerospike nozzle, both printed as single copper-alloy parts. (leap71.com) An aerospike nozzle is the strange one. Instead of one bell-shaped exhaust tube, it uses a center spike so the exhaust can adapt better as outside air pressure changes with altitude, which is why engineers have chased it for decades even though it is hard to cool and hard to manufacture. (leap71.com) The hidden enabler is additive manufacturing, better known as 3D printing. It lets engineers build cooling channels, injectors, and other internal shapes as one metal piece, which cuts down the number of welded joints and makes fast iteration physically possible. (sheffield.ac.uk; leap71.com) Solar cells solve a different engineering problem. Instead of turning hot gas into motion, they turn sunlight into electricity, and their main limit is how much of the light spectrum a single material can absorb before the rest is wasted as heat or passes through. (nature.com) A tandem solar cell stacks two light absorbers like two fishing nets with different mesh sizes. The top layer catches higher-energy light first, and the lower layer catches wavelengths the top layer misses, which is how tandem designs can beat the ceiling of a single-layer cell. (nature.com) The Lingnan University-linked result sits in that race. A January 2026 paper in Nature Communications reported a perovskite-organic tandem solar cell with 27.11% power-conversion efficiency, with a certified value of 26.3%, using a buried 2D/3D interface designed to cut defect losses and improve charge extraction. (nature.com) Perovskite is the promising but temperamental material here. It can be tuned to absorb the right colors for the top cell, but wide-bandgap perovskites often lose voltage through microscopic defects and can suffer stability problems, so the paper focused on cleaning up that buried interface where charges often get stuck. (nature.com; techxplore.com) Put the two stories together and the pattern is the same: faster engineering loops. In rockets, the loop is software-to-print-to-hot-fire in weeks; in solar, the loop is interface tweaks that push efficiency to 27.11% while improving stability, which is how hardware gets better one test stand and one lab stack at a time. (leap71.com; nature.com)