LLM Foundations | Guide DAO

Starts today

11 modules

Course

LLM Foundations

Understand LLMs from the mechanism to the application: tokens, probabilities, attention, training, API requests, context, evals, and security. Every module ends with JavaScript practice in the browser.

Private Discord community

Practice, quizzes, and a final project

Mentors and checkpoints

Lifetime access and updates

About the courseThe course works from the inside out. In the first half, you build a language model in the browser, step by step, that writes chess openings. It learns from 2,385 opening lines from the open Lichess database. You measure every improvement — BPE tokens, temperature, longer context, gradient-descent training — with numbers: how many games are legal and how many the model simply copied from the book. In the second half, you build a press-office assistant for a chess team: a client for an OpenAI-compatible API, document search with citations, an answer eval, and prompt-injection defenses. In the capstone, the assistant withstands 30 attacks without revealing a player's final preparation. You need basic JavaScript and school-level math. You do not need to play chess or have an API key for the main track.

Explain from the mechanism why a model hallucinates, cannot count letters, and answers the same question differently
Build a chess-opening generator in JavaScript: a BPE tokenizer, a temperature sampler, attention, and gradient-descent training
Write a client for an OpenAI-compatible API with tools, document search, and an answer eval with confidence intervals
Defend an assistant against prompt injection: the secret bypasses the model, and code checks permissions

LLM Foundations

Who the course is for

New

Developers who know JavaScript and want to understand how LLMs work inside

Pro

People building on LLM APIs who need to reason about cost, context, and model errors

Engineers implementing RAG, structured output, and evals

Technical leads who choose models and verify them with their own evals

Newcomers to machine learning: school-level math and basic JavaScript are enough

Have questions?

We’ll reply within 15 minutes

Or contact

a consultant directly

Curriculum

Eleven modules lead from the model's mechanism to applications built on it: tokens, probabilities, attention, and training, then APIs, context, evals, and security. Every module ends with JavaScript practice in a browser editor and a short quiz.

How an LLM Writes Text

Tokens

Choosing the Next Token

Inside the Transformer

How a Model Learns

Learning Platform

Short lessons, practice, and final artifacts live on one platform. Wallet login does not require a separate password.

Wallet Login
Connect MetaMask from any device and resume where you stopped. You do not need a separate password.
Short Lessons
Lessons are split into small blocks. Complete one in a short session, then resume from the same point later.
Practice Throughout
Assignments apply the material right after a lesson. In larger programs, separate outputs build toward a final project or capstone.
One Workspace
Lessons, quizzes, assignments, and course materials stay together. Technical tasks can use the editor and syntax highlighting.

CommunityEach block ends with an assignment. Keep the intermediate results and assemble the final piece step by step.

Reviews and Community

Practice Reviews

Mentors review code, models, product decisions, and research artifacts in a format that fits each program.

Expert AMAs

Research Briefings

What You Leave With

Finish the program with a verifiable result you can show to a team or use in your next project.

Final Project

Build the program's main result against clear criteria and take it to a state another person can verify.

Portfolio of Work

Keep intermediate artifacts, sources, and decisions so others can see how you worked and why you reached the result.

Next Step

Record the result's limits, open questions, and next actions for your own project, a team, or a new role.