01 / 05LANGUAGE
Large Language Models
Stop treating language models like magic.
From text pieces and attention to learning, alignment, and dependable language-model systems.
- Foundations
- Architecture
- Pre-Training
- + 4 more frontiers
Five courses · one connected field guideUnderstand intelligent systems.
Understand intelligent systems.
From first principles.
Build a durable mental model of the ideas behind language, imagination, generation, decisions, and embodied action—then test that understanding in the browser.
Explore all courses No account required. Learn at your own pace.
- Courses
- 5
- Connected lessons
- 182
- Guided hours
- 78+
THENEURALFIELD
01 · LANGUAGELLMsText → tokens → representations → predictions02 · IMAGINATIONWorld ModelsObservation + action → possible futures03 · CREATIONGenerative ModelsRandomness + structure → controlled samples04 · DECISIONSRL & ControlExperience + reward → better decisions05 · ACTIONEmbodied AIPerception + language → safe physical actionInteractive course map · select a field to enter
Five ways into the machinery.
Each course is complete on its own. Together they form a map of how modern intelligent systems represent, predict, create, decide, and act.
02 / 05IMAGINATION
World Models
Learn how machines imagine before they act.
From observations, actions, and possible futures to imagination, planning, robotics, and bounded deployment.
- Foundations
- Representations
- Learning Dynamics
- + 4 more frontiers
03 / 05CREATION
Generative Models
Turn randomness into controlled creation.
From familiar randomness and sampling choices to latent models, diffusion, control, and original experiments.
- Probability to Generation
- Autoregressive & Latent Models
- Flows & Energy Models
- + 3 more frontiers
04 / 05DECISIONS
Reinforcement Learning & Control
Teach systems to improve through consequences.
From action, consequence, and partial observation to deep, model-based, offline, safe, and original agent experiments.
- Sequential Decisions
- Values & Temporal Difference
- Deep Value Learning
- + 4 more frontiers
05 / 05ACTION
Embodied AI
Bring intelligence out of the screen.
From sensor-to-action tasks and partial observation to demonstrations, closed-loop control, and original embodied-system experiments.
- Embodied Problems
- Perception & State
- Demonstrations & Data
- + 3 more frontiers
Start with the question you cannot stop asking.
You do not need to take the courses in a fixed order. Choose the system you want to understand, then follow the cross-course links when your curiosity expands.
Large Language Models
Stop treating language models like magic.
From text pieces and attention to learning, alignment, and dependable language-model systems.
44 lessons17 hours
Enter the course Understanding is something you do.
Every lesson moves from orientation to mechanism, from mechanism to action, and from action to evidence that your understanding transfers.
- 01→
Orient
Know the outcome, prerequisites, and why the idea matters.
- 02→
Learn
See the plain-language intuition and the precise mechanism.
- 03→
Try
Change an input, trace the state, and explain what moved.
- 04→
Test
Commit to a changed-case answer and get specific feedback.
- 05
Extend
Connect the idea to primary sources and a larger build.
See it. Change it. Explain it.
Invisible mechanisms become worked traces, inspectable diagrams, controlled simulations, and changed-case assessments. A click is never mistaken for mastery.
CHANGEOBSERVEEXPLAIN
Small teaching fixtures make state and causality visible without a paid API.
Required explanations, practice, checks, hints, and answers work without an account.
Deterministic checks diagnose the choice and offer a concrete retry path.
Build cumulative artifacts with observable criteria, failure logs, and worked references.
Your next question is the trailheadThe machines are complex.
The machines are complex.
Your path in does not have to be.
Choose one field, begin with one concrete mechanism, and leave with a mental model you can use to build, evaluate, and challenge intelligent systems.
Choose your course