Prediction comes before execution
The student writes a hypothesis before running code or simulation. The task becomes a test of a cause-and-effect model, not a hunt for a memorized answer.
The OmniLab method
Perturbation-based learning
Perturbation-based learning starts from a working system. A student changes one thing, predicts the effect, and compares that prediction with the observed divergence.
The mutation is controlled
The change is small and inspectable: an operator, a line of code, or a physics-world parameter. Attention stays on the cause instead of noise from many changes.
Divergence is the learning moment
OmniLab shows the original and mutated worlds side by side. The Socratic AI coach asks questions until the student connects the observation with theory.
Looppredict -> mutate -> diverge -> reconcile
Domainscode and 2D physics
Approachexplanation before answer
Frequently asked questions
How does a perturbation cycle work in OmniLab?
Each cycle has four steps: the student predicts the effect of a change, applies one controlled mutation, observes how the mutated world diverges from the original, and reconciles the difference with theory.
Why does prediction come before execution?
Writing a hypothesis before running the code or simulation turns the task into a test of a cause-and-effect model instead of a hunt for a memorized answer.
What kinds of systems can students perturb?
OmniLab supports working Python and JavaScript code and 2D Newtonian physics simulations. Every mutation is small and inspectable — an operator, a line of code, or a physics parameter.
What happens when a student gets stuck?
The Socratic AI coach asks guiding questions until the student connects the observed divergence with the underlying theory — it does not hand over the answer.