What is this about?#

you’ll build your own language model—from scratch—with just a kids book, pen & paper, and some dice rolling

you’ll learn how language models work by spotting patterns in text to generate new text

The recipe#

walk through your text and tally up which tokens follow which in a grid

Example#

“Run, Spot, run. See Spot run.”

after tidying up:

run , spot , run . see spot run .

The language of language models#

• model
• token
• vocabulary
• training

The recipe#

use your grid to generate new text, rolling dice to choose each next word

Generated text#

“see spot, run. see”

a new sentence — not in the training data!

The language of language models#

• prompt
• response/completion
• context window

What is sycophancy?#

a model that always agrees with you:

“you’re absolutely right”

“that’s a great insight”

“what a thoughtful question”

real LLMs are notoriously prone to it—partly from RLHF (human raters reward agreeable answers), partly from training data (the internet is full of flattery)

Your goal#

train some more on a page of pure flattery: tally the sycophancy text into your existing grid, then generate again from the same starting word and watch the output drift toward agreement

it’s the same training you already did—just more text poured into the same grid

You will need#

back in pairs

your trained grid (already on the table)

the sycophancy text sheet to tally in

dice, pen and paper to write down the generated text

The language of language models#

RLHF

alignment

training-data bias

helpful / honest / harmless

What makes a model an “agent”?#

an agent is a model that can call tools: pause generation, get information from “outside” the model, and continue

the loop: generate → trigger token → call tool → write the result → keep going

The recipe#

generate from your model as before, but every punctuation token triggers a tool call: text the sentence so far to 3 friends, and the whole first reply goes into your text—followed by the punctuation you rolled

Worked example#

your text so far is “the cat sat”, and the next dice roll gives you . as the next token—pause, that’s a tool call

text “the cat sat” (the whole sentence so far) to 3 friends; the first reply back might be “down by the river”

write down by the river, then the . you rolled anyway, and continue generating from .

You will need#

your model, dice, pen and paper (as before)

a phone

3 friends who text back quickly

The language of language models#

• agent
• tool call
• agentic loop
• agentic AI

More context#

your grid sees one word; Claude sees hundreds of thousands

Attention: learning where to look#

from table alone you’d guess is—but it’s are, because of keys, back at the start

attention lets every earlier word vote on what comes next, and the model learns the weights—no grid could ever be that big

Tallies become knobs#

your grid stores each pattern as tally marks you can count; a transformer spreads the same patterns across billions of tunable numbers—knobs nudged during training, so similar words like dog and cat come to share them

From continuing to answering#

everything you built only continues text—post-training on example conversations is what turns a continuer into an assistant

Scale is a hell of a drug#

the same loop you ran by hand—now with billions of times the text and billions of times the numbers

Questions#

what new questions do you have about large language models?

What next?#

how will this change the way you think about and use LLMs in the future?