When we published Building LLMs for Production in 2024, we thought we had written a book we could keep updating as the field evolved.
We were wrong.
Not about where the field was going, but about how much the engineering itself would change. In 2024, much of the work was still about getting more out of the LLM.
Two years later, the model is only one component of the system.
The field is more mature now, and so is the problem we wanted the book to solve. Rather than anchor another book too closely to today’s stack, we wanted to focus on the engineering problems that better models alone will not solve.
So what began as an update became book number two:
AI Engineering for Production, launching October 20.
The question behind it is simple: once the model is already good, what actually gets a system the rest of the way to production?
We also want to finish this book a little differently. Rather than show up on October 20 with a finished book, over the next few weeks, we will share our progress as we make the final passes, and we would genuinely like your help making it better.
If there is a problem you think we have missed, something you want us to go deeper on, or a question you would want the book to answer, reply to any of those updates. We will be reading them as we fine-tune the final version and, wherever it makes the book stronger, working your feedback in. Then, on launch day, we are bringing everyone together. We will open up the book, show you the AI engineering stack we actually use today and the decisions behind it, and spend a big part of the session answering your technical questions live.
Follow the book + join us on October 20
What’s left to engineer when the model is good? was originally published in Towards AI on Medium, where people are continuing the conversation by highlighting and responding to this story.