Book Wrap Up

Dear reader, I have been paid for “AI work” (for many interpretations of what that even means) since 1982. I certainly find LLMs to be the most exciting tool for moving the field of AI further and faster than anything else that I have used in the last 43 years.

I am also keenly interested in privacy and open source so I must admit a strong bias towards using open source software, open weight LLMs, and also systems and infrastructure like Ollama that enable me to control my own data. The content of this book is tailored to my own interests but I hope that I have, dear reader, covered many of your interests also.

In the last example in the previous chapter I “pulled a fast one” in that I didn’t use a local model running with Ollama. Instead I used what is the most powerful commercial LLM Claude Sonnet 3.5 because it generates better code than any model that I can run on my Mac with 32B of consolidated memory using Ollama. In my work, I balance my personal desires for data privacy and control over the software and hardware I use, with practical compromises like using the state of the art models running on massive cloud compute resources.