Apple has published eight videos from its 2025 AI Reasoning and Planning Workshop, a two-day event held last July featuring presentations from Apple engineers and academic researchers. The newly released talks cover critical areas including embodied AI agents, visual reasoning robustness, and internet-scale agent training.

The workshop brought together experts from UCLA, Stanford, UC Berkeley, and other institutions to discuss nearly 30 studies on AI reasoning and planning. Presentations included topics like adaptive parallel reasoning, reinforcement learning for interactive agents, and benchmarking agent capabilities across diverse domains. The delay in publishing these videos—seven months after the event—highlights how quickly AI technology evolves, yet the content remains valuable for understanding current research directions.

Despite the time lag, these videos offer insights into how Apple's engineers and academic partners are approaching the state of the art in AI reasoning and agentic systems. The talks reveal both the progress made and the open challenges shaping future developments in foundation models and autonomous agents. For anyone tracking AI advancement, these presentations provide a window into the technical thinking driving the next generation of intelligent systems.