TL;DR- Read “Vibe coding” – Ignore the title, read the book.

There’s an old saying, if someone told you to jump off a bridge, would you? Given one’s need for self-preservation, the expected answer is: No. But given Gene Kim’s history of writing persuasive and engaging technical content, I think that would be hard to say no to.

Gene Kim wrote the famous book “The Phoenix Project: A Novel About IT, DevOps, and Helping Your Business Win.” It’s the kind of book that you find lying around the office (well that’s how I found out about it), and then pick it up, and you end of reading it out of curiosity. It tells a story about the technical workplace, burnout, walking away, and then coming back with a new windfall that helps your team improve and remove stressful processes. It’s not so much of a prescriptive technical reference as much as it is a motivational convincer to work with DevOps. It’s about how bringing in new technologies and structure into an environment can bring massive improvements. It’s hard not to relate and like books like that.

Well, Mr. Kim teamed up with Steve Yegge and wrote a book on AI Assisted coding. The book is called “Vibe Coding.” I resisted reading this book due to the name and the negative connotation of generating code without a little foresight to the risks that comes with it. I picked up the book after hearing more about “Gas Town” and listening to his podcast interview on the Programmatic Engineer podcast. The book takes a casual approach at learning to start with AI Assisted Development. More on the critque later.

Well, AI is very unlike most other technical topics that you’d see on the Manning publication catalog. AI/LLMs and coding isn’t something you can learn concepts, apply it and then implement it at once. Models and their engines change, and your output is highly variable based on training, context, and your prompt. Many introductions into AI development and assisted tools tends to be too academic in the introduction (talking about tokenization, activation energy, and neural networks), and they lean less on selling the usage of the tool. Personally, I believe that teaching AI tools needs to acknowledge the trust issue first and then help to ramp up your experience slowly.

Back to the review

The book takes an approach where it leads you to some motivational chapters and narrative story building. Then it pushes you to do some light examples with an LLM. Even after getting experience with LLMs, I found the examples to be surprising and engaging. (It asks for you to generate a bouncing ball animation on a web page, and there’s another example of creating a Flappy Bird clone). It helps to build motivation on the subject and introduces the concept of FAAFO when working with these tools. (Fast, Ambitious, Fun, and Optionality) The book operates on an undertone of getting back to the joy of building things.

Throughout the book, the book breezes over topics such as context windows/context engineering, RAGs, context saturation, why and how MCPs are needed (A topic that I feel is very poorly explained in external resources), and it also reviews the overall view of how the development cycle changes when using AI Assisted tools. Later in the book it goes into classifying the process and decisions that goes into making software and how AI assisted development modifies or replaces or enhances those previous processes.

Lastly, what did I really enjoy about this book? The book is empathetic towards individuals who are approaching AI (Well LLMs) with hesitation. It’s also a strong reminder that coding should be fun again, and not focused on worrying about extreme edge cases in stressful situations. Additionally, it guides you through the journey into becoming an AI Assisted Developer (from chat oriented programming all the way to full AI Agent orchestration).

What do you read after this book?

Take a look at Chip Huyen’s book: “AI engineering : building applications with foundation models.” This is a great book to follow up with because it provides technical deep dives with content you can put into use quickly.