Researchers at the University of Pennsylvania have developed tiny, programmable robots capable of leaping meters into the air, spinning, and autonomously planting seeds. Published in the journal Science, the study led by Shu Yang and Yaoye Hong reimagines the knot not as a passive fastener, but as an active energy storage system. The robots consist of a fiber thinner than a millimeter, featuring a Kevlar core for strength and a surrounding shell of liquid crystal elastomer (LCE) for flexibility. These fibers are twisted and knotted to store elastic energy. When exposed to heat between 60 and 90 degrees Celsius, the LCE shell contracts, loosening the knot trigger and releasing the stored energy in a fraction of a second. This mechanism allows a millimeter-long knot to jump nearly two meters, a height hundreds of times its own size. By altering the knot's topology, researchers can program specific movements, such as flipping with an overhand knot or spinning with a figure-eight knot. To enhance control during flight and landing, the team attached maple seed-inspired wings to the fibers. These wings stabilize the robot and direct its descent. The system is specifically designed to penetrate soil for agricultural purposes. Unlike previous bio-inspired seed carriers that relied on rainfall to activate, which can be inconsistent or damaging, this new system uses heat from sunlight. The jumping force generates penetration pressure approximately thirty times greater than rain-activated methods, ensuring seeds are securely buried for germination. Early experiments with pine and arugula seeds confirmed successful growth after being planted by the robots. While this technology offers promising applications for autonomous reforestation and agriculture in diverse environments, the development of such small, adaptive machines represents only one step toward broader goals. The current design uses well-understood materials to study underlying physics, with future iterations aiming for environmentally friendly components and lower activation temperatures. Separately, the landscape of scientific research is shifting with the emergence of autoresearch, where AI models conduct, validate, and iterate on their own experiments. In 2026, Anthropic released a notable study on superalignment, a field dedicated to ensuring superhuman AI remains interpretable and controllable. This work addresses critical questions about AI behavior and safety, aiming to solve the challenge of controlling advanced systems. However, the success of such autonomous research raises concerns about potential inequality in access to AI capabilities.