INNOVATIONResearch-grade; demonstrated on locomotion (DrEureka).90% confidence

The 10,001st World

Train across 10,000 randomized simulations and reality becomes just the 10,001st — sim-to-real by distribution, not fine-tuning.

Problem it solves

The sim-to-real transfer gap that normally requires expert hand-tuning per parameter.

Best for

Sim-to-real transfer in robotics; reasoning about generalization across realities.

Not ideal for

Tasks where the real-world distribution can't be parameterized or randomized in sim.

Overview

Why this framework exists

Fan's reframing of domain randomization. Train a policy across 10,000 parallel simulations, each with slightly different physics (gravity, friction, weight). An agent that masters all 10,000 configurations treats the real physical world as just the 10,001st sample from the same distribution — so it generalizes zero-shot, no fine-tuning. DrEureka demonstrated it: a robot dog learned to balance and walk on a yoga ball purely in sim, then transferred to the real world untouched. The deeper claim: virtual and physical are 'different realities on a single axis,' not different problems.

Core principles

4 total
  1. Randomize physics across thousands of parallel sims, not one careful sim.
  2. Master all N configurations and reality is just configuration N+1.
  3. This yields zero-shot sim-to-real without further fine-tuning.
  4. Virtual and physical are points on one reality axis, not separate domains.

Origin story

How this framework came to be

Given in the 'Is the virtual world in the service of the physical world?' section, grounded in the DrEureka follow-up to Eureka.

Source

Traced to primary
Source · PODCAST
Jim Fan on Nvidia's Embodied AI Lab and Jensen Huang's Prediction that All Robots will be Autonomous
Sequoia Capital (Training Data) · 2024
Open source →

Related frameworks

Browse all Innovation →