Researchers at the University of Toronto Engineering have utilized an AI-driven, self-sustaining laboratory to discover six novel 3D-printable metal alloys engineered for extreme high-temperature environments. Led by Canada Research Chair Yu Zou and Ph.D. student Ajay Talbot, the project demonstrates a closed-loop discovery platform capable of accelerating materials science innovation for aerospace and energy sectors. Conventional high-performance alloys, typically nickel- or cobalt-based with numerous minor additives, are reaching their performance limits in applications like jet engines and nuclear steam generators. Identifying viable alternatives across tens of thousands of compositional permutations is traditionally hindered by data scarcity and experimental bottlenecks. To overcome these constraints, the Toronto team implemented an active learning framework that integrates machine learning, computational modeling, and robotic fabrication. The system strategically selects limited material samples for manufacturing and thermal testing, continuously ingesting experimental data to guide subsequent iterations without requiring extensive pre-existing datasets. Operating entirely autonomously, the self-driving lab focused on compositionally complex nickel-cobalt-chromium systems. Within weeks, the algorithm identified six promising formulations. Two specific alloys demonstrate marked advantages over current industry standards, particularly Inconel 625. The first formulation, composed of 12 percent nickel, 62 percent cobalt, and 26 percent chromium, maintains superior hardness under extreme thermal stress up to 600 degrees Celsius, outperforming Inconel 625 by 4.5 percent in laboratory trials. A second variant, containing 36 percent nickel, 14 percent cobalt, and 50 percent chromium, exhibits exceptional oxidation resistance at temperatures reaching 1,000 degrees Celsius, exceeding the benchmark material by 85 percent. These capabilities directly address critical failure points in aerospace propulsion and power generation, where extreme heat rapidly degrades conventional metals. The research also emphasizes the structural advantages of additive manufacturing. Because the alloys are optimized for layer-by-layer fabrication, engineers can design components with graded material properties, varying density and hardness within a single part to achieve optimal strength-to-weight ratios. This capability is particularly valuable for manufacturing intricate geometries that traditional machining cannot produce. Published in npj Advanced Manufacturing, the findings validate the efficacy of data-lean active learning models in uncharted materials design spaces. While the current study utilizes a three-element system as a proof of concept, the research team plans to expand the platform complexity to ten or twelve elements, targeting operational thresholds of 1,200 degrees Celsius. The project received partial support from the University of Toronto Acceleration Consortium, which promotes cross-sector collaboration to accelerate AI-driven scientific discovery. This closed-loop methodology establishes a scalable pathway for rapid materials innovation, potentially transforming how advanced alloys are developed for next-generation high-stress engineering applications.