1From Words to Numbers: Tokens and EmbeddingsA language model can't read letters — it only does math. Here's how your words become numbers it can work with.AI foundations 2Attention: How a Model Decides What MattersIn any sentence, some words matter more than others. Attention is how a model figures out which ones — and it's the heart of how these models work.AI foundations 3Writing One Word at a TimeA language model never writes a whole answer at once. It predicts a single next word, adds it, and does the whole thing again — over and over.AI foundations 4The Memory Trick: How Models Remember the ConversationWriting one word at a time should mean re-reading the whole chat every single word. A clever shortcut called the KV cache stops that waste.AI foundations 5Why Long Chats Get Slower and PricierLength, speed, and cost are all tied together. Once you see the two kinds of work a model does, the whole bill makes sense.AI foundations