An update about my book “Building effective LLM-based applications with Semantic Kernel”

In January, I started writing my second book. I wanted to have another go as an author on a topic close to me: making Large Language Models practically usable for people with a technical focus. So, I started writing the book “Building effective-LLM applications with Semantic Kernel.” Five months into writing the content and publishing chapters, it’s time for some reflection and an update on the plans going forward.

What’s my book about again?

For those who haven’t read it, my new book, “Building effective LLM-based applications with Semantic Kernel,” covers how to build AI applications that use an LLM for things other than chat. I also cover some chat topics, but to be honest, that wasn’t my main goal for the book. For me, the value is in automating content tasks as part of existing solutions. You can read my book at Leanpub: https://leanpub.com/effective-llm-applications-with-semantic-kernel/

The current state of the book

I’m writing the final pieces of chapter 10, and I have some editing to do before I can send it to the readers.

The first two months of the writing process flew by faster than the past three months because I had more time in January and February to work on the book. I couldn’t do my normal work back then, and now the conference season has started, and I’m back doing my regular work again. Of course, I am happy I can do my regular work, but it also means things have slowed down.

Overall, the book is well received. I keep getting positive comments from people in the hallway at work and sometimes online about my book and how it helps people learn to use Semantic Kernel. Some people gave me constructive feedback about layout issues and unclear samples that I wouldn’t have caught myself because I’m neck-deep into writing content.

I’m happy people send me feedback, both positive and negative. It’s nice to see people care enough to let me know how they perceive the book. It helps me do my best work and finish the job.

According to my original plan, I have three more chapters to write, and finishing the work is getting harder. I’m hitting writer’s block more often now, and the combination of work and writing isn’t ideal. On the one hand, I want to write the final chapters and start improving the earlier chapters. But at the same time, I’m not overly motivated now in general. I struggle with finding the right flow. I’ve made some changes I wanted to inform you about.

Going forward

With the feedback in my pocket and the fact that I’m not as motivated now, I spend some time thinking about the book’s final portion.

I will restructure the final chapters, so I have a complete book but don’t have to spend another 4 months writing the final chapters.

The new structure for the final chapters will be:

• Chapter 10: This chapter currently covers how to use the process framework in Semantic Kernel and how to use it to build intelligent request routing workflows. I will expand this chapter with some information about artist-critic workflows as a variation on the topic.

• Chapter 11: Agents, this chapter will contain the content for chapters 13 and 14. I’ll cover combining tools and prompts into agents and using multiple agents to solve the more complex problems in LLM-based applications.

I’m removing the original chapter 12 because it feels redundant now that I have most of the chapters ready. The original chapter 11 would have been nice because I feel that artist-critic workflows are an interesting topic to cover, but I think you can build one of those workflows with the information in Chapter 10. Finally, I condensed chapters 13 and 14 into one chapter for my sanity.

All these changes mean I can complete the book sooner. That means I can start improving existing content sooner and finish up the Python samples. It also means I can move on to marketing the book and discussing the content at conferences.

Here’s my call to action for you

Again, I’m happy with the number of people reading the book. I sold 59 copies of the book without marketing it anywhere, and over 400 people downloaded the content from the GitHub repository. The download figures tell me that the book is useful.

If you are wondering whether you should jump on the LLM bandwagon, I recommend reading my book. Even if you’re skeptical, I can convince you that it’s a valuable tool to have in your toolbox. And yes, I’m not asking you to use LLMs everywhere — it’s a bad idea!

You can find my book up on Leanpub: https://leanpub.com/effective-llm-applications-with-semantic-kernel/

Read my book and let me know what you think. I’m happy to clarify things and help you gain experience building chatbots, workflows, or even agents.

And thanks to everyone who has already read my book. I don’t care if I sell thousands of copies. I care about helping people understand.