2.0
Datasets
Behavioral and perceptual datasets for frontier AI labs and applied AI teams. Built where data collection is hardest: agentic web interactions, ego-centric video, and embodied spatial tasks.
Commerce Agent Traces
End-to-end recordings of AI agents navigating e-commerce: product search, comparison, cart, and checkout flows across live retail environments. Built by Machine Commerce for training and evaluating commercial agents. In development.
Ego-Centric Activity Data
First-person video and sensor streams capturing object interaction, spatial navigation, and manipulation tasks. Designed to train robots and spatial AI systems on how humans move through and act on the world. First release planned for early 2027.
CartBench
A structured evaluation suite for commerce agents, built from Machine Commerce behavioral traces. Covers product discovery, price comparison, cart management, and end-to-end checkout, scored for task completion, step efficiency, and robustness to real-world failure modes. In design; planned for early 2027.
Retail Shelf Observations
Annotated shelf photos with per-facing product labels, price detections, and inventory counts, being assembled across retail categories and store formats. Planned to be paired with ground-truth POS data where partners provide it, enabling training of models that connect visual shelf state to commercial outcomes.