---
title: "How a language model is grown"
description: "The seven steps by which engineers produce an LLM, showing why nobody designs what ends up inside."
canonical: https://tomsharman.com/bookshelf/if-anyone-builds-it-everyone-dies/how-a-language-model-is-grown/
updated: 2026-11-03
---

# How a language model is grown

> The seven steps by which engineers produce an LLM, showing why nobody designs what ends up inside.

Eliezer Yudkowsky and Nate Soares's framework, from If Anyone Builds It, Everyone Dies, pp. 21 to 101.

Framework 1 of 5 from [If Anyone Builds It, Everyone Dies](https://tomsharman.com/bookshelf/if-anyone-builds-it-everyone-dies/index.md) by Eliezer Yudkowsky and Nate Soares. Category: AI.

## The steps

1. **Render the input as numbers.** Turn a text fragment into a series of numbers.
2. **Provide parameters.** Acquire storage for billions of numbers, each slot a parameter.
3. **Fill with random weights.** Fill the slots with randomly chosen numbers called weights.
4. **Choose the architecture.** Set the rules that combine the input with the weights.
5. **Read the output.** Treat the output numbers as probabilities for the next letter or token.
6. **Train by gradient descent.** For every weight, work out how a tiny change would make the answer more correct, nudge it that way and repeat over trillions of words.
7. **Generate text.** Feed each predicted token back in to continue the text, then tune the model to answer as an assistant.

## Where it comes from

From chapter: PART I: NONHUMAN MINDS, pp. 21–101

## The animation

24 seconds, 4:5, no sound.

- MP4: https://tomsharman-downloads-006143767029.s3.eu-north-1.amazonaws.com/downloads/bookshelf/if-anyone-builds-it-everyone-dies/how-a-language-model-is-grown.mp4
- MP4, dark: https://tomsharman-downloads-006143767029.s3.eu-north-1.amazonaws.com/downloads/bookshelf/if-anyone-builds-it-everyone-dies/how-a-language-model-is-grown-dark.mp4
- Still: https://tomsharman-downloads-006143767029.s3.eu-north-1.amazonaws.com/downloads/bookshelf/if-anyone-builds-it-everyone-dies/how-a-language-model-is-grown.png
- Still, dark: https://tomsharman-downloads-006143767029.s3.eu-north-1.amazonaws.com/downloads/bookshelf/if-anyone-builds-it-everyone-dies/how-a-language-model-is-grown-dark.png