Prompts to Processors

A Live Book by Gaurav Ramesh

Prompts to Processors

Just enough AI to make sense of it all.

What this book is

This is a conceptual guide to AI that covers the entire stack — from what happens when you type your first prompt, through the transformer architecture and large language models, down to the GPUs, data centers, and semiconductor physics that make it all possible.

It's a distillation of my learning and insights as I make sense of the landscape.

It is not a tutorial. There are no code walkthroughs, no toy projects, no framework quickstarts, and no math! Instead, each chapter builds a mental model that helps piece together the bigger picture.

What it prepares you for: architecture reviews, making sense of industry debates and business shifts, technical trade-offs, spotting overclaims, sifting signal from noise on the internet, and knowing when something doesn't add up.

Who this is for

It is primarily aimed at people with a background in software engineering, computer science, or a related field—who may not have a background in AI. But the content varies from introductory to advanced, so it suits a wide range of readers.

It's also for anyone who wants to understand AI but doesn't know where to start; for anyone trying to make sense of AI conversations but feels lost in the noise; and for anyone who wants to grasp not just ever-changing trends, but underlying principles and architecture.

And for anyone wondering where they might fit in the AI stack in the next era of software engineering.

Feedback and suggestions are welcome. You can reach me at promptstoprocessors@gmail.com.

Table of contents

I love Early Release books from O'Reilly that provide raw, unedited content from upcoming books, allowing members to read and learn new technologies months before official publication. This is written in the same spirit and published as it's written.

It'll contain roughly 24 chapters (this might change as I write) across five parts, going from top of the stack to the bottom. Because I'm writing as I learn, in the early hours of the morning before my kids are up, I expect that it'll take 2-3 weeks for each chapter to be written and published.

I also plan to write the chapters in a semi-random order based on whatever I feel like learning the most about at the moment.

Part I — Artificial Intelligence

  1. 1. Intelligence, Artificial and Otherwise Planned
  2. 2. The Machine Learning Mindset Planned
  3. 3. Data — The Part Engineers Underestimate Planned
  4. 4. How Learning Actually Happens Planned

Part II — Large Language Models

  1. 5. Language as a Prediction Problem Draft
  2. 6. The Transformer Architecture Planned
  3. 7. Pre-training, Fine-tuning, and Alignment Planned
  4. 8. Reasoning, Memory, and the Limits of LLMs Planned
  5. 9. The Competitive Landscape Planned
  6. 10. Engineering with LLMs — Prompt, Context, and Harness Engineering Planned

Part III — Mathematical & Software Foundations

  1. 11. The Mathematics of Representation Planned
  2. 12. Optimization Theory and Practice Planned
  3. 13. The Software Stack Planned
  4. 14. Training at Scale Planned

Part IV — The Hardware Stack

  1. 15. Why AI Demands Different Hardware Planned
  2. 16. GPUs — Architecture and Dominance Planned
  3. 17. Custom Silicon — TPUs, NPUs, and Accelerator Design Planned
  4. 18. Interconnects, Networking, and Cluster Architecture Planned
  5. 19. Data Centers and the Infrastructure of AI Planned
  6. 20. How Chips Are Designed Planned
  7. 21. How Chips Are Made — Semiconductor Manufacturing Planned

Part V — The Bigger Picture

  1. 22. AI Safety and Alignment — A Technical Introduction Planned
  2. 23. AI and the Engineering Profession Planned
  3. 24. Where the Field Is Going Planned

Follow the book as it's written.

Get notified when a new chapter is ready.

No noise, just the work. Unsubscribe anytime.

Share this

X LinkedIn Hacker News Bluesky