Generative AI Fundamentals Map

Map/Physical world/Wave 5

Embodied agents

General agents that act in the physical world, where every mistake has a physical cost.

stageEmerging Open on the map →

Facts

Lane: Physical world
Wave: 5 of 12 (build order, not a date)
Needs: 10 boxes all the way back through hard links
Unlocks: 4 boxes downstream through hard links
Leverage: 0.8 points · stands in the way of 4 of 53 stages · rank 12 of 57
Cluster: Robots and the physical economy
When: Here since 2025

Evidence as of October 2026

π0.5 tidies homes it never saw in training. Simulation is "essential but insufficient", so real action data remains the scarce input.

Watch
Tasks completed in unseen homes without a remote operator.

Timeline note
First general robot policies working in unseen homes, 2025.

01Physical Intelligence: π0.5

Around it

Needs · 7

Multimodal perceptioncapability · Built today
Robot data and foundation modelscapability · Emerging
Long-horizon agentsstage · Built today · helps, not required
World models and simulationcapability · Emerging · helps, not required
Rollbackcapability · Emerging · helps, not required
Rare-earth permanent magnet supplycapability · Built today · helps, not required
Precision robot actuatorscapability · Emerging · helps, not required

Unlocks · 5

Home robotsstage · Speculative
Long-latency civilisationstage · Speculative
Explosive economic growthstage · Speculative
Self-replicating robot factoriesstage · Speculative
Cognitive abundancestage · Speculative · helps, not required

Brakes

Works against this box

China critical-mineral export controlsbrake · Built today
US ban on foreign-made robots and dronesbrake · Built today

Loops and off-ramps

Physical data flywheelreinforcing loop
Capex pullbackoff-ramp · cut off
Taiwan chip disruptionoff-ramp · cut off
Power shortfalloff-ramp · cut off
Capability plateauoff-ramp · slowed
Training data ruled outoff-ramp · slowed

Openings researchers' judgement, October 2026 scouting

Startup room 32/100 · rank 106 of 124 Research room 52/100 · rank 53 of 124

Crowding5 Lab overlap5 Small-team fit1.5 Buyer today2 Research openness4

Depth: deep
Players: Figure ($39bn); Skild ($14bn+); PI; NEURA (~$7bn); Apptronik (~$5.3bn); NVIDIA GR00T; DeepMind Gemini Robotics
Wedge: Physical-AI safety and assurance layer: independent real-world eval fleets, runtime failure detection and certification-oriented testing (weakly evidenced, no startup found). startup