Research
Ten cross-cutting docs synthesizing the state of Physical AI in 2026 — AV industry, robotics foundation models, simulation, labeling, world models, open problems, and a sequenced learning roadmap.
- 00
Overview — Physical AI in 2026
~30 minCross-cutting synthesis of the field: data is the bottleneck, the data engine is the product, the modular/E2E pendulum is dissolving.
- 01
AV industry and data
~90 minWho collects what at what scale: Tesla, Waymo, Mobileye, Wayve, Waabi. Fleet logs vs customer-shadow vs sim vs world models.
- 02
Robotics foundation models
~60 minThe VLA recipe — RT-X, OpenVLA, π0/π0.5, Helix, Gemini Robotics. Open X-Embodiment, demo collection economics.
- 03
Simulation and synthetic data
~90 minThe competitor map. Applied Intuition Simian, Nvidia Cosmos + Isaac, CARLA, Waymax, the death of pure-rendering shops.
- 04
Labeling and data curation
~120 minThe most important doc for the role. Data-engine philosophy, FiftyOne/SAM2/OpenSCENARIO tools, foundation models as labelers.
- 05
World models and generative
~60 minCosmos, GAIA, GR00T-Dreams, Wayve PRISM-1. World models as data engine and as eval substrate.
- 06
Open problems and benchmarks
~60 minThe RAND 11-billion-mile result; SOTIF, ODD, V&V; benchmarks that drive the field forward.
- 07
Learning roadmap with mini-projects
~45 minThe 8-to-12-week plan. Each phase produces a tangible artifact; each project has a definite "done" criterion.
- 08
Connections and gaps
~30 minHow the projects link together — the critical chain, the round-2 audit, the bijective folder-rename mapping.
- 09
Research frontier and outlook
~50 minWhat's next: open research questions, where the next 12–24 months of progress will land, what to watch.