A new AI planning algorithm, VNT-PA, improves navigation in complex environments by using spatial context indexed by camera pose, outperforming traditional methods in training efficiency and navigation performance. This development is relevant to people building and operating AI agents, particularly those working with visual navigation and planning.