This book shows you how to build your own large language model from scratch. Not just use tools like ChatGPT, but actually understand and create the technology behind them.

It’s written for curious beginners: you don’t need a CS degree, advanced math, or prior programming experience—just a laptop, Google Colab, and the patience to follow along step by step.

The book is organized as one continuous, hands-on project:

• Foundations – what AI and LLMs really are, why they work, and where their limits are.
• Python Crash Course – just enough Python and essential libraries to understand and run real models.
• Build Your First LLM – implement tokenization, embeddings, attention, and the full Transformer architecture.
• Make It Useful – fine-tuning, prompt design, and building practical applications on top of your model.
• Ship It – options for deployment and running your model in the real world.

Every concept is explained in plain language, with analogies instead of jargon, and every line of code is walked through in detail. You’ll run existing models in minutes, then gradually replace the black box with pieces you built yourself.

All code is designed to run in Google Colab, with a dedicated notebook for each chapter linked directly inside the book. The notebooks are also available on GitHub at https://github.com/FirstLLM/code. To help you check your understanding, the book includes example answers to the review questions in an appendix.

⚠️ Early access notice: This is an in-progress book. Parts I and II are largely complete, and later chapters are being added and refined over time (including recent updates to Chapters 5–7). Buyers get all future updates for free.