| Horatius' Last Stand | Is there a better way to go? | |
| Jonathan Livingston Seagull | no limits eh? | |
| Machines of Loving Grace | "A country of geniuses in a datacenter" | |
| Situational Awareness | Things are gonna get interesting with the next few OOMs… | |
| The Logintaka | a dialogue between a Guru and a Newbie | |
| How To Become A Hacker | Sounds like a good group of folks | |
| The Bitter Lesson | Make search and learning limited by compute, then increase compute. | |
| Self-Verification, The Key to AI | Experience is bitterest, but it is the best way to learn | |
| Poetry and Ambition | McPoems? shit | |
| Hanuman Chalisa | Jai Shri Hanuman! | |
| The Anabasis of Alexander | That's a pretty good reply, given the context | |
| Mozart, Writing Symphonies and Young Lawyers with Questions | The master is just the student which is furthest along! | |
| Occam's Razor | simpler! | |
| Gunga Din | You're a better man than I… | |
| I love my wife. My wife is dead. | You will probably want a drink for this "fine-man"'s letter. | |
| Be Drunk | Wine, Poetry or Virtue? | |
| A Mathematician's Apology | A peek into the mind of a mathematician | |
| Electric Meat | Grab a few things! | |
| On Bullshit | Its despicable! | |
| A Message to Garcia | Are you capable of carrying the letter? | |
| You and Your Research | There IS difference between "Good" and "Great" | |
| The Goddess of Everything Else | Wouldn't it be based if we did some of that? | |
| Kindness to Kin | Reject fitness score, embrace cripplemind | |
| 500 Million, Not a Single One More | get vaxxed! | |
| A Big Little Idea called Legibility | An Authoritarian High-Modernist Recipe for Failure | |
| The Glass is already Broken | Enjoy it! | |
| The Laws of Trading – Review | You're always trading! | |
| Good Research and Good Papers | What constitutes a good paper? | |
| The Techno-Optimist Manifesto | ai ~= alchemy, we ought work with it to build Nuclear Fusion. | |
| Research Debt | Debt exists everywhere. | |
| How to Do Research (MIT AI Lab) | A fantastic collection of timeless, ad-hoc research advices | |
| The Maintenance Race | Skilled maintainers advise never trying to solve a new or complex problem without a thorough mulling first. | |
| Jibun Kaigi | A titular meeting of the selves! | |
| Tips for Writing Technical Papers | by Jennifer Widom? it should be Wisdom! | |
| Notes from The Pragmatic Scientist | Useful items to keep in mind | |
| The Ten Commandments of Egoless Programming | It is always good to refer back to this, especially when wheels have been spun. | |
| Principles of Effective Research | Hamming's essay provides the raison d'etere, this provides the practical | |
| The OPT-175B Logbook | What a fun time! | |
| The Art of Science and Engineering | This really ought to be required reading for anyone looking to do something useful in Science and/or Engineering, If nothing else read Chapters 18-through-20 on Simulations, quite a useful perspective for the way things seem to be going. | |
| Vox Populi | "In these democratic days…" | |
| How Complex Systems Fail | They tend to fail in non-trivial & intertwined manners | |
| Markdown Syntax | A take on markdown from the creator himself, John Guber | |
| Conscious intention to redirect attention | FOCUS | |
| A Recipe for Training Neural Networks - Karpathy | Timeless! | |
| A DL Tuning Playbook | So simple, so sane. | |
| EleutherAI's cookbook | Worth gander mate! | |
| The Art of Debugging | Art is correct-a-mundo | |
| Art in the Sciences of The Artificial | "Sometimes algorithms can be odes to certain aspects of nature rather than faithful reproductions of them. And so it goes with art." | |
| Debugging Backwards in Time | "If you've got a snake's tail and you pull on it long enough, you will get to its head." | |
| Magic Ink | "An artist who paints with magic ink." | |
| How to Write a Git Commit Message | If for no one, do it for your future self! | |
| A Note About Git Commit Messages | For your future self, for the love of God. | |
| Epigrams on Programming | If you program, you are in good company. | |
| Raw Performance Ranking of GPUs | GPUs really do go BRRRRRR | |
| Slow Science | Take the time to do the thing right, le journey is le reward. | |
| File over app | A compelling case to use obsidian | |
| Inside Cuba's D.I.Y. Internet Revolution | No es fácil. | |
| The Hardware Lottery | What you will discover is determined in part by the legos you have to build with. | |
| Write Like A Programmer | Writing is coding in a natural language | |
| Tadao Ando | Don't you just love concrete? | |
| The Origin of the Research University | "18th century German bureaucrats didn’t invent the idea of making new contributions to human knowledge. Their great contribution was to institutionalize it…" | |
| On the Design of Text Editors | Why can't all software be as functional and ELEGANT as this? | |
| Welcome to the Era of Experience | In some sense, the difference between simulation and experience is the program from which you get your observations and rewards – a process on a computer, or a process on reality. | |
| The future of AI is open-ended | This is where things appear to be heading, or at least, ought to be. | |
| 10 Bulls | Have a read | |
| Master Foo and the Recruiter | Its a new floor, very complex! | |
| Amateurs and Professionals | Where do you get your reward, from percieving the form or from executing the function? | |
| You Already Understand! | "I hit you! Ha ha ha ha!" | |
| In Praise of Shadows - Tanizaki | It is no coincidence that darkness and "日本" go hand in hand, it is a well-thought out relationship. | |
| All AI Models Might Be The Same - Morris | If all of these models are modeling reality, and there is only one reality, are we then just training "one true model"? | |
| Equinox: neural networks in JAX via callable PyTrees and filtered transformations | It seems that wonderful things tend to come from thinking thoroughly about something and then sketching it out. | |
| Tensor Logic: The Language of AI | Ya know, everything exists on some N-dimensional tensor… | |
| Lánczos interpolation explained | Now this is a good example of explaining something | |
| Stringly-typed: it's not that simple! | "type-safety theatre" | |
| The Adolescense of Technology | "Here we go!" mario voice | |
| 店舗オーナーを経たからこそ、できること | The Latte Tokyo closed (sad), these are the owner's "思い" | |
| ml-engineering by stas00 | A based chap's repo with nice ml learnings + utils (nccl interconnect tests of the highest quality) | |
| Why AGI Will Not Happen by Tim Dettmers | So the science fiction is reality thing, as of world state in Dec 2025, maybe not | |
| The Harness is the Backend - iii | That is one clean abstraction, open source as well… | |
| Twelve Virtues of Rationality | Its a nice read, accompany a coffee | |
| Why Japanese companies do so many different things | "Most importantly, Toto has a very lucrative sideline in the fabrication of memory chips." – Japan is the best | |