Every lesson, in one place

No math, no hype. A clear picture of what today's AI systems are, what they're doing when they answer you, and why that matters for how you use them.

Prediction, tokens, and training — the three ideas that explain almost everything a chatbot does. Enough depth to be useful, not so much that it stops being interesting.

Hallucination isn't a bug that will be patched next month. Here's why it happens, which questions trigger it most, and the habits that catch it before it costs you.

Most bad AI output comes from a vague ask. Five rules that turn a one-line question into a prompt that gets you something you can actually use.

The model only knows what you put in front of it. How to choose what to include, what to leave out, and why more context isn't automatically better.

Role, examples, step-by-step, and critique. Four reusable shapes that cover most of what you'll ever need to ask, with a worked example of each.

An honest comparison of the major assistants for real tasks — writing, research, code and analysis — including where each one is genuinely weaker.

AI is a brilliant research assistant and a terrible source. A workflow that gets you the speed without inheriting the confident nonsense.

Why AI drafts read the way they do, and how to use the model for the parts it's good at — structure, alternatives, editing — while the voice stays yours.

Build something small and genuinely useful end to end — a recurring task that runs itself. The best way to understand AI is to make it do a job for you.

Agent is the most oversold word in AI right now. What the term actually describes, what these systems can genuinely do today, and what they still can't.