Start Here: A Map of The Computist Journal
Last updated: 2026-06-01.
If you found this newsletter through a link, a recommendation, or by accident, welcome. This post is the map — a guided tour of what I write about, why, and where to start.
What this is about
For five years I’ve been writing about computers — what they can do, what they can’t, what they’re made of, and what they mean. The newsletter started as a side project for sharing things I thought were beautiful and ended up as the place where I think out loud about everything that matters to me professionally.
I keep coming back to one belief: the public conversation about AI is broken. It oscillates between hype and dread, with very little in between. Neither extreme is grounded in what computers actually are. The path through is to look at the machinery — algorithms, computation theory, formal systems, the long history of what we’ve learned about what can be computed and what cannot. The map underneath the territory.
That’s what this newsletter is for. It’s also why the writing here ranges from “what’s the fastest sorting algorithm” to “is the universe a computer.” Those questions look unrelated. They’re not. They’re the same question at different scales — and the more you know about one, the more sense the others make.
If you read enough of what I write, you’ll find I’m an opinionated technologist with a soft spot for foundational ideas, a deep suspicion of bullshit, and a stubborn belief that understanding the inside makes you better at navigating the outside. If that sounds like you, you’ll like it here.
Five paths through the archive
The newsletter is loosely organized into five sections. Pick the one closest to what brought you here; each ships roughly two posts a month.
🤖 If you want to understand what AI really is — and isn’t
This is the largest section, and the one I think about hardest. The thesis: large language models are extraordinary at some things and structurally limited at others, and most of the confusion in public AI discourse comes from not knowing which is which. Three places to start:
The Insurmountable Problem of Formal Reasoning in LLMs — Or why LLMs still can’t, and probably never will be able to fully reason. If you only read one post on this newsletter, it should be this one.
Why Reliable AI Requires a Paradigm Shift — Hallucinations aren’t a bug we’ll patch in the next release. They’re the price of the architecture we’ve chosen, and the only way out is to change the architecture.
Artificial Neural Networks Are Nothing Like Brains — Busting the biggest public misconception in AI. The metaphor is so wrong it actively blocks understanding.
These three together form a kind of mini-syllabus on the real limits of modern AI. They’re also the spine of my book Mostly Harmless AI — more on that below.
🧠 If you want to see why computation itself is beautiful
I came to AI through computer science, and I think the field has lost something by skipping the foundations. The deepest, most satisfying ideas in computing aren’t about models or tools — they’re about what can be computed and what can’t. This section is my attempt to make those ideas accessible without dumbing them down.
The Most Beautiful Algorithms Ever Designed — A very opinionated short list. Read this and you’ll never look at “an algorithm” the same way again.
The Power of Abstraction — How to think like a Computer Scientist. The one mental move that separates programming from computer science.
The Answer to All Questions — How the quest to answer all mathematical questions became the origin story for a new science. The Hilbert-Gödel-Turing arc, told as drama instead of textbook.
If those land, the section has dozens more, from the P vs NP problem to formal language theory to the foundations of machine learning.
💻 If you want to build something
The Mondays of this newsletter are conceptual. The Thursdays are practical. This section is where I show working code, walk through implementation decisions, and occasionally release something into the world.
Chat with your PDF — A simple streamlit app using LLMs and vector databases to answer questions from a PDF. The minimum viable RAG, demystified.
How I Built the Database of my Dreams — A pure-Python vector database for prototyping. Build AI apps 100× faster, with one dependency.
Realtime 3D in Pure Python + Numpy — A side project that went way too far. The kind of thing you write because you want to know if you can.
This section also doubles as the trailer for whatever I’m currently building — most of the open-source projects on my GitHub started as a Thursday post.
✒️ If you want to argue about the bigger stakes
Sometimes I step out of the technical voice and write an opinionated essay. Most of these are about how AI is reshaping science, education, and the way we work — and what to do about it. None of them have a clean answer; that’s the point.
The Techno-Pragmatist Manifesto — A level-headed response to techno-pessimists and techno-optimists. Where I plant my flag.
Teaching the Mostly Harmless Way - Part I — Principles, practices, and tools I use in all my teaching endeavors. Half autobiography, half pedagogical argument.
The Three Minds of a Computer Scientist — The Scientist, the Engineer, and the Hacker. A useful taxonomy for anyone in this field.
If you came here from a more humanities-flavored corner of the internet, this section is probably where I sound most like the rest of the writers you read.
💡 If you want to chase the big questions
The deepest section, and the one I’m most personally invested in. Questions about minds, machines, science, and reality — through the lens of computation. No clean conclusions; just careful thinking.
Can Machines Think? — What Alan Turing’s seminal paper Computing Machinery and Intelligence actually says, and what it doesn’t.
Is Science the Answer to Life, the Universe, and Everything? — Part I — Exploring the world of scientific theory and practice. Where epistemology meets engineering.
Are Brains and Computers Alike? — The boldest of all theories of mind. A careful walk through the computational theory of mind and its critics.
This section moves slowly — maybe one post every other month — but the conversations in the comments are the best on the whole newsletter.
The books
Across these five paths, certain arguments come back often enough that they grow into books. Right now there are five, in various states:
Mostly Harmless AI — the finished one. A book-length argument about what AI can and can’t do, distilled from the Mostly Harmless AI section. Just shipped v2.0.
The Algorithm Codex — a growing pay-what-you-want collection of essays on the most beautiful algorithms ever designed. Always in motion.
The Science of Computation — early access. The book version of the Science of Computation section: foundations, computability, complexity, the whole stack.
The Hitchhiker’s Guide to Graphs — early access. Graphs as the universal data structure, and what you can do with them once you take them seriously.
How to Train Your Chatbot — early access. The practical companion to Mostly Harmless AI — building real LLM applications without the magic-thinking.
If you want them all in one place, the Computist Compendium bundles everything at a discount.
The code
On top of writing, I maintain a long string of open-source projects: AI tools, web frameworks, Python libraries, and assorted cool toys. Some of them have been quietly running in production at companies I’ve never heard of; others are weekend experiments that ran away from me. They all live on my GitHub.
The pattern is usually the same: a Thursday post explains why I built the thing, the repo has the code, and a month later it shows up in a Monday essay as a worked example. If you’re following the build track, watch both feeds.
What you get if you subscribe
Every article on this newsletter — every essay, every tutorial, every deep dive — is free, and will be free forever. I had the incredible fortune of receiving a world-class education at no cost. Writing publicly is my humble way of paying some of that back.
Starting June 22, 2026 there will also be a paid subscription, for $7/month or $70/year. Paid subscribers will get:
Every book I’ve ever published, free, forever. A 100%-off code refreshed continuously, covering everything in the catalog above and everything I publish from this point onward. At current prices that’s well over $200 worth of material, and it grows every month.
Book drafts as PDFs, often months before public release — half-edited, sometimes rough, always the latest thinking.
A weekly Friday digest with what shipped, what’s coming next week, and an open AMA thread.
Prioritized attention in comments, chat, and Office Hours.
Founding members will lock all of it in for life. If you’re a professional engineer or researcher, you can almost certainly expense the subscription to your employer — I’ll be happy to help you draft the email.
What’s next
Every month, a new chapter of a book or a new open-source release. Every Sunday, a short This Week at The Computist Journal note on what shipped and what’s coming. And every once in a while, an editorial when something in the field is too important to ignore.
This map gets updated as the archive grows. If you bookmark one post on this newsletter, make it this one — it’ll always point you at the current best version of where to start.
The subscribe button is at the bottom of every post on this site, and the one at the top of this one too. If you’ve read this far, you already know whether this is your kind of newsletter. Hit it, and I’ll see you in the next post.
— Alejandro