Where to go from here
What you now know, what’s still open, and a capstone challenge.
Look how far you’ve come
You started with the number line. Along the way you learned to program, then met algebra, logarithms, trigonometry, vectors and matrices, eigenvectors, derivatives and integrals, probability, likelihood and entropy. Then you built — yourself, in code — linear and logistic regression, k-nearest neighbours, decision trees, k-means, neural networks, an autograd engine, a convolutional network, word2vec, recurrent networks, attention and a working GPT. Finally you looked at the frontier: assistants, fine-tuning, retrieval, reinforcement learning, diffusion, multimodal models, efficiency, reasoning, evaluation and safety. Along the way you finished nine projects of your own.
The key ideas are the ones you’ve met here. The current frontier is mostly these same ideas, scaled up and refined.