Favorite Books and Papers of 2025

In 2025 I tried to spend less time on social media and more time reading long-form content. I read 19 books which isn't a lot, but it's a lot for me. I also read a lot of papers, but didn't track or remember all of them. I switched to libro.fm for all my audio books so I'll link to that since it helped me de-Amazon my life. If you want the full list of what I read subscribe to my newsletter. I'll be sending the full list with summaries soon. Here's my favorite long-form content from 2025. The Invention of Air A book about scientific discoveries in the late 1700s. This book had a line that I've thought about all year and how it applies to other things in life. >When our ability to measure something improves by an order of magnitude our understanding of that thing fundementally changes. This idea applies to so many things, and it was the most insightful idea to me. Material World How sand, salt, iron, oil, copper, and lithium changed the world. I loved this review of how these materials affect our modern world and where they come from. I paired nicely with Chip War which was good, but not one of my favories. The Cathedral and the Bazaar This is a collection of classic papers written by Eric S. Raymond that I had never read. It's freely available online--I bought a physical copy. Most of the papers were written more than 20-30 years ago and they apply to today's and tomorrow's software and open source ecosystem. My most memorable quote had to do with how to create quality software and I think this directly applies to the world of AI. >Enjoyment predicts quality. If you like the software you're writing, it's going to be better software. ChatGPT is Bullshit This is a 2024 paper I read in 2025 and it was great! It sussinctly explained a lot of the things I struggled to put into words. The tl;dr >...lying and hallucinating require some concern with the truth of their statements, whereas LLMs are simply not designed to accurately represent the way the world is, but rather to give the impression that this is what they’re doing. >Because these programs cannot themselves be concerned with truth, and because they are designed to produce text that looks truth-apt without any actual concern for truth,it seems appropriate to call their outputs bullshit.