What is this thing?

Start from nothing. Finish knowing what a language model is, what it cannot do, and why.

12 concepts, about 72 minutes of reading at roughly six minutes each. Free to read, with no account.

What is in this track

What the thing actually is — A large language model is one very large machine for guessing what text comes next. (free)
One piece at a time — The model writes a small piece of text, reads its own output, and writes one more. Hundreds of times. (free)
The pieces are not letters — The model never sees letters or words. It sees tokens: chunks of text drawn from a fixed list. (free)
Who decides where to cut — The tokenizer is a frozen list of chunks, learned by merging the most common pairs of characters over and over. (free)
The prompt is the page — The prompt is everything handed to the model before it writes. Nothing off the page exists. (free)
The instructions you never see — Every message carries a label saying who said it, and one of them was written by the company, not you. (free)
The context window — There is a hard limit on how much text can sit in front of the model at once, and what happens when you exceed it depends on who is asking. (free)
It has no memory — Nothing carries over between messages. The whole conversation is re-sent and re-read every single turn. (free)
Training, then running — Training is the enormous one-off process that set the numbers. Inference is what happens each time you press send. (free)
It stopped reading at some point — A model’s training text ends on a date. After that date it knows nothing, and it cannot tell. (free)
Why it makes things up — Confident invention is not a glitch bolted onto the machine. It is the same mechanism that makes it work. (free)
There is not one model — There are many models from many makers, they change constantly, and the differences are big enough to matter. (free)

After this: Turning the dials

Every track · Pricing · Claims we checked and could not stand behind