How LLMs Work

Interactive visual guide based on Karpathy's lecture

What is How LLMs Work?

How LLMs Work is an interactive visual guide that explains how large language models function under the hood. Based on Andrej Karpathy's popular lecture, it breaks down the complex mechanics of LLMs through interactive diagrams and demonstrations rather than text alone. The tool helps you understand concepts like tokenisation, attention mechanisms, and how models generate text one token at a time. It's designed for people wanting to move beyond surface-level knowledge of AI systems, whether you're a student, developer, or someone curious about how these models actually work. The visual approach makes abstract concepts concrete, letting you experiment and see results in real time.

Key Features

Interactive visualisations of LLM architecture and how tokens flow through the model

Step-by-step breakdown of the transformer architecture components

Demonstrations showing how language models generate predictions

Code examples and explanations for key concepts

Browser-based interface requiring no installation or setup

Pros & Cons

Advantages

Free to access with no account required
Visual learning approach makes complex topics easier to grasp
Based on well-respected educational material from Andrej Karpathy
Practical examples help bridge theory and real-world model behaviour

Limitations

Best suited for people with some technical background or interest in machine learning
Limited scope; focuses specifically on understanding model mechanics rather than building or using models

Use Cases

Learning how transformer models work before diving into implementation

Understanding what happens inside an LLM when you prompt it

Teaching others about LLM fundamentals with visual aids

Preparing for technical interviews about AI and machine learning

Satisfying curiosity about why LLMs behave the way they do