Field notes
Dataset profiles, annotation methods, Physical AI research, practical Studio guidance and our monthly view of the landscape—written to support real decisions.
Latest analysis
119 articles · January–August 2026
A decision-focused review of world models, reward models and rollout tooling: what the evidence can support, where teams can over-read it, and how to validate the next move.
A decision-focused review of measured annotator calibration: what the evidence can support, where teams can over-read it, and how to validate the next move.
A decision-focused review of actionable failure classification: what the evidence can support, where teams can over-read it, and how to validate the next move.
A decision-focused review of multiview 3D review: what the evidence can support, where teams can over-read it, and how to validate the next move.
A decision-focused review of closed-loop planning scenes and metrics: what the evidence can support, where teams can over-read it, and how to validate the next move.
A decision-focused review of closed-loop failure recollection: what the evidence can support, where teams can over-read it, and how to validate the next move.
Our July outlook on why data work is moving from commodity label volume toward model-specific evaluation, multimodal evidence and accountable operations.
A decision-focused review of release acceptance tests: what the evidence can support, where teams can over-read it, and how to validate the next move.
A decision-focused review of simulated trajectories in challenging conditions: what the evidence can support, where teams can over-read it, and how to validate the next move.
A decision-focused review of slice-based embodied evaluation: what the evidence can support, where teams can over-read it, and how to validate the next move.
Research archive