Start with the right questions
Build intuition for AI, data, predictions, and responsible experiments.
Go from your first line of Python to your first working AI system. Clear explanations, real code, and room to experiment.
FROM FIRST PRINCIPLES TO FIRST PROJECTfor idea in your_curiosity:
understand(idea)
experiment(idea)
build_something_real()Follow the path in order, or find the concept you need today.
36 lessons · Every lesson includes theory, Python code, a practice task, and a knowledge check.
Build intuition for AI, data, predictions, and responsible experiments.
Write small, reliable programs and turn messy inputs into useful structures.
Work with vectors, probability, loss, and gradient descent through tiny experiments.
Train, evaluate, and compare models without hiding what happens underneath.
Build neurons, calculate gradients, and train a small network from scratch.
Explore tokenization, retrieval, language models, and attention.
Design grounded answers, bounded tools, useful evaluations, and efficient systems.
Examine privacy, fairness, prompt injection, monitoring, and model documentation.
Complete four runnable Python projects with evaluation and extension challenges.
What happens when a model becomes more selective? Move the slider and see the answer. Our interactive labs make abstract ideas tangible.
Explore the full lessonSix invented predictions, one decision boundary. A higher score is classified as positive when it meets the threshold.
No. Start with Foundations, then Python. Later modules build on that knowledge. Every example is explained, and every practice task includes a solution to compare with your own approach.
Python 3.11 or newer, a text editor, and a terminal on your own computer. All 36 examples use the standard library. No GPU, paid API, or package downloads are required for the lessons.
The interactive concept labs run directly in your browser. Python examples have copy and download controls for local execution; the site does not run uploaded code on a server.
Completion is stored locally in this browser, without an account. Clearing site data removes it, and it does not automatically sync to another device. If browser storage is blocked, progress lasts only for the current page session.
A support-message classifier, a study-note search engine, a neural network that learns XOR, and a local prediction API backed by SQLite. These are complete educational projects with explicit limits and extension exercises.
Original lessons and examples by Artificials. Synthetic datasets are labeled throughout. Optional primary references are provided for deeper study; estimated times include the exercises.