Skip to content

Learn from first principles

Mathematics, machine learning, and everyday decisions. Start with an example, work through the calculation, then read the general rule.

Probability → learning from data. Describe randomness, fit predictions, and evaluate them on new observations.

Slopes, areas, and gradients, with the reasoning behind each rule. Starts from school algebra.

Turn a table of measurements into predictions. Follow vectors, matrix products, and projections through one small dataset.

The Math overview connects the foundations. Backpropagation follows a prediction’s error backward to the parameters that produced it.

One dataset, from fitting to a final test connects averages, squared error, gradients, and model selection. Work with eight observations and see why a perfect training fit can lose on new rows.

Game theory: where your effort goes is a short essay on incentives, competition, and choosing what is worth pursuing. The worked market examples are optional.

Options: reading the Greeks uses one option quote to explain delta, gamma, theta, vega, and rho. Change one input and see what happens to the price.

Transformers follows a next-token prediction through the model. LLMs in the Wild covers data, prompting, evaluation, and failure. The AI overview connects these to backpropagation, post-training, speech, and images.

Research papers, from first principles unpacks AI papers with diagrams, calculations, and checks for understanding. The DeepSeek-V4.1-Flash walkthrough connects attention, experts, memory, and recurrence to current model designs.

Try a calculation before opening its answer. Predict what will change in a diagram, then move the control and check. If a formula looks unfamiliar, identify what is fixed, what varies, and what you need to find.

Definitions open where you’re reading. Your place and practice history are saved on this device.

Definition

Read the full glossary entry →