Learn everything
From a question to a course you approve
Say what you want to understand
Write the goal in your own words, with what you already know and the language you want to study in.
Unfurl researches the subject
It searches for sources, reads them, and shows in the conversation what it read and why.
Review and approve the program
Change the outcomes, the order or the scope. Nothing is built until you approve a version.
Lessons and practice are built and checked
Each section gets reading material and activities, and every activity is validated before you start.
Understanding AI,
from idea to application
Find your foundations
Connect the essential ideas in data, probability, and Python. Begin with what you know and give unfamiliar ideas time to settle.
Data, probability, Python
Understand how models learn
Train a small model, investigate what it learns, and test how well that understanding travels to new examples.
Training, evaluation, generalization
Build an AI system
Bring models, instructions, tools, and useful information together. Follow a request through the system and see where each decision belongs.
Models, tools, retrieval
Make something of your own
Choose a problem you care about. Build a working project, explain its limitations, and decide what you would improve next.
A project and a reasoned explanation
Learn through experiments
and explanations
Adjust a model, connect the parts of a process, and explain the choices you make.
Change a model
See how it fits
Choose how much freedom a model has. Compare its shape against the same eight data points.
Connect the parts
Follow the process
An AI system is more than a model. Explore how a question, useful information, and a response belong together.
Choose a part to look closer.
Be specific about what the system needs to help someone do.
Make a choice
Explain your reasoning
Consider the evidence behind a choice and the conditions where it might fail.
What did you notice that supports your choice?
29 kinds of practice, chosen for what the subject needs
- Recall reconstruction
- Confusable discrimination
- Rule induction
- Procedure reconstruction
- State tracing
- Outcome prediction
- Cause diagnosis
- Self-explanation
- Experimental investigation
- Causal intervention
- Source investigation
- Relational mapping
- Analogy transfer
- Representation translation
- Estimation
- Example construction
- Counterexample construction
- Plan construction
- Artifact construction
- Diagnosis and repair
- Adversarial testing
- Constrained optimization
- Argument and rebuttal
- Critique and revision
- Misconception tutoring
- Information gap
- Role simulation
- Audience mediation
- Strategy selection
One conversation
Several courses
Revise a course or create another in the same conversation. Each course has its own program; your library brings your courses and learning progress together.