The manufacturer demonstrates walking, sitting, kicking, grasping, skating and recovery. Use these as starting points for observation; the projects below are café learning ideas, not additional shipped features.
Project 1: the movement detective
Choose one official demonstration and watch it twice. First follow the feet; then follow the head. Write down which movements appear coordinated and which details you cannot determine from the camera angle. This exercise turns passive viewing into a question about control. It works before you own any hardware and makes your next documentation search more specific.
Project 2: a repeatability notebook
With an appropriate setup and the manufacturer’s guidance, repeat one supported behavior from a consistent starting condition. Record attempts, successes and failure descriptions. Keep the surface and command the same. A small table of repeated observations is more informative than one spectacular clip. Do not intentionally damage the robot or assume recovery makes arbitrary falls harmless.
Project 3: compare two virtual conditions
In a documented simulator or training environment, compare a baseline with one available setting changed. Predict the effect before running it. Then compare what happened with your prediction. If the result surprises you, inspect the assumptions before making a second change. This is an experiment-design exercise, not a claim that the browser sandbox exposes every training parameter.
Project 4: explain a failure generously
Take one failed attempt and describe it without calling the robot “stupid.” Was the input ambiguous? Was the initial posture different? Did the environment violate an assumption? Translating a failure into observable conditions is a useful robotics skill. Share a clear description alongside the successful result, and distinguish an idea for a new behavior from a policy that has actually been trained and tested.
Go to the source
Reviewed September 7, 2026. Prices, availability and software can change.