What is this?
The main Academy teaches you to work with AI. The Expert Library teaches you how the AI itself works — the knowledge that lets you pick the right model, read a model card without blinking, and understand exactly what you're running (and where). Read the six pages in order, or jump to whatever you need. No account, no quiz — this is a reference shelf.
Parameters & Model Sizes
What a parameter actually is, what the "B" in 8B means, and which sizes run on which hardware.
Start HereOpen Weights & Model Families
What "open weights" really means (and doesn't), the major model families, and the license fine print.
Explore Open WeightsModel Types
Chat models, reasoning models, embedding models, vision models — what each one is for, and when to use which.
Learn the TypesQuantization
How a 70-billion-parameter model shrinks to fit modest hardware — and what the compression costs you.
Understand QuantsChoosing a Model
Tokens and context windows, how to read a model card, why benchmarks lie a little, and a decision table that matches tasks to model sizes. The capstone — everything else builds to this.
Read the CapstoneLong-Horizon Work
How agents carry real projects across hours and days: why long tasks fail (drift, context exhaustion, credit assignment), the five-pattern scaffold that fixes it, and how "time horizon" gets measured. The bridge from model mechanics to finished work.
Learn Long-Horizon Patterns