Core Automation investigates alternatives to conventional model scaling, exploring new learning algorithms and neural network architectures that extend beyond traditional pretraining and reinforcement learning. Its work includes building AI agents and systems designed to perform iterative tasks, optimize workflows, and support research activities.