Private course preview

Robot Learning,
gently.

A 12-lesson path from policies and imitation learning to a working ANIMA-Kiwi creature—grounded in ETH Zürich’s 2026 course, its guest spotlights, and a practical LeKiwi build sequence.

12guided lessons
21reviewed ETH videos
4runnable ETH tutorials
4Kiwi build gates

One loop, increasing power

Take the lessons in order on your first pass. Each one adds only the machinery needed for the next. Watch the selected lecture slice, read the three-idea model, do one trace, then complete the lab or transfer prompt.

On-ramp

Build the loop before learning the algorithms.

01
See the policy loopTranslate a robot task into observations, actions, a policy, and—only when useful—an MDP.
45 min

Learning behavior

Learn first from people, then from consequences.

02
Learn from demonstrationsStart with the simplest useful robot-learning recipe, then repair its two characteristic failures.
60 min
03
Learn from consequencesBuild enough reinforcement learning to reason about algorithms without turning the lesson into an equation wall.
75 min

Action representations

Choose how a policy represents motion and time.

04
Generate actions, don’t average themUnderstand why robot actions are often distributions—and choose an action head by quality and latency.
65 min
05
Turn behavior into sequencesUse the transformer lens without pretending robot data is just text.
55 min

Foundation models

Connect world models, VLAs, and reasoning without losing the robotics.

06
Imagine before actingBuild a useful mental model of world models: compress, predict, and improve behavior in imagination.
65 min
07
Build a generalist policyUnderstand the modern VLA recipe and the systems decisions hidden behind the model diagram.
70 min
08
Reason, evaluate, and stay skepticalPlace embodied reasoning inside the control loop and learn to separate a compelling demo from a supported claim.
60 min

Build ANIMA-Kiwi

Turn the theory into a measured, safe, end-to-end creature demo.

09
Kiwi lab 1 — Make trustworthy dataBring up the LeKiwi body and create the smallest dataset worth training on.
90 min
10
Kiwi lab 2 — Train the action brainTrain in an order that keeps failures interpretable, then deploy with a measurable action buffer.
120 min
11
Kiwi lab 3 — Add memory, voice, and expressionKeep the creature responsive while perception, planning, and policy inference run at different speeds.
110 min
12
Kiwi lab 4 — Close the creature loopIntegrate the three tracks with state machines, safety gates, and a demo runbook you can actually rehearse.
120 min