Preface

This monograph is a work in progress—and an academic experiment on building in public: I am making its content available online as I write it, in both PDF and HTML formats. I’d be delighted to have you follow along, and welcome your feedback. Information on how to contact me can be found on the book’s website. There is also a mailing list you can subscribe to in order to learn when major updates are released, which will never be used for any other purpose. If you prefer not to receive email, please see the main website for other options.

Naturally, as this book is a work in progress, it might contain errors—both technical and typographical. If you find one, please contact me immediately by filing a GitHub issue. Effort has and will be made to ensure publicly-visible content has reached a certain level of maturity and can be trusted. The current table of contents contains the plan for the book’s content, but at present, many chapters have yet to be written. They are marked with obvious placeholder text that indicates they are under construction, and will appear in future drafts.

This book comprises the academic side of what I am currently working on, but it is not my sole focus—or, in some sense, even my primary focus. As such, progress is likely to be uneven and will appear at random times. If you’d like to know about what else is on my mind, please visit my personal website to learn more. And, if you’re particularly excited about a planned chapter’s contents, please feel free to let me know, so that I have a better sense of what to prioritize.

Once this book reaches a complete draft, I plan to rewrite this preface. The new one will likely become more personal, with reflections on my career—after all, I did make the unconventional decision to leave academia in favor of other ambitions while writing this book, and nonetheless plan to actually get it done.

I originally got interested in Bayesian decision-making because I thought studying it deeply could give clues for how to develop artificial general intelligence—which, in the form of large language models, now exists. Today, I see Bayesian decision-making as a promising way to study how language models actually work, and in particular how they handle exploration. I hope this book inspires the next generation to further connect these research areas. Finally, I thank my wife Kamilė for her love and support, which has made this work possible.

Alexander Terenin
Vilnius, Lithuania, August 2026