What should I write next?
Asking for your feedback!
Mostly Harmless AI is almost out of the alpha stage! I have a couple of chapters left to refactor, and then I’ll send it to my awesome editor for absolute obliteration.
In the meantime, I’ll shift my focus to finishing, or at least moving forward significantly, another of my in-progress projects. So I want to ask for your feedback on what you think I should write next.
Here are the options.
How to Train your Chatbot is a hands-on, tutorial-based book on building applications with generative AI, especially LLMs. Contrary to Mostly Harmless AI, this one is targeted specifically at software developers, to give you the tools to build chatbots, coding assistants, customer service agents, and even RPG dungeon masters, using today’s most advanced techniques and industry best practices. And it’s pure code from page 1.
The Hitchhiker’s Guide to Graphs is mix of math and algorithms, a book focused on graph theory and applications. Every chapter introduces a new algorithm or technique on graphs—from basic pathfinding to network analysis to graph inference)—with intuitive explanations and optional deep-dive mathematical sections for those interested in the underlying theory.
The Science of Computation is my most ambitious project. It’s a full book on all things computer science, from foundational theory to how microchips works to the internet to videogames to artificial intelligence. It’s the book I wish I could have read in high school. But it’s so big in scope, that my purpose would be targeting only Part 1 for now, the foundational theory of computer science (computability, formal languages, and complexity).
So these are the options. I want to focus the next two to three months in finishing the first draft of one of these three projects. Which one will depend heavily on what you would prefer to read next.
As usual, all drafts will appear as articles in this blog, free to read online for everyone interested.
Ready? Time to vote!
PS: In the meantime, I’ll keep posting the remaining finished drafts of Mostly Harmless AI, because your feedback has been invaluable so far.
I'm working on a system to valuate "negative truths" i.e. opinions, lies, mis- and disinformation campaigns, that find fertile soil with large groups within societies. This chapter may be interesting for you to look at.