There is no posttraining data commons
The LLM revolution ran on a pretraining commons that humanity built for other reasons. Posttraining consumes tasks, not documents, and no such commons exists. If the open models community wants...
Every blog post that took off on Hacker News, ranked by the votes of its best submission. 5,964 posts.
The LLM revolution ran on a pretraining commons that humanity built for other reasons. Posttraining consumes tasks, not documents, and no such commons exists. If the open models community wants...
Who is the Assistant? We investigate the character that most modern language models inhabit when interacting with users.
A blog about software craft and quality
My Happiest, Most Grounded Distraction
In the Game of Life, which still lifes can be produced by crashing gliders together? We’ve known for a few years that the answer cannot be “all of them”, because in 2022 Ilkka Törmä and Ville Salo found a patch of still life that, if it exists in the universe, must have existed since the beginning of time. And so…
Update 2026-04-07 Update: 2026-04-11 Update: 2026-04-17 Let’s kick the developers out! In the ongoing saga of The Document Foundation (TDF), their Membership Committee has decided to eject from membership all Collabora staff and partners. That includes over thirty people who have contributed faithfully to LibreOffice…
Look, I really didn’t want to like Duet. I don’t want to be an Anthropic fanboy and it annoys me when they do things better than the competition. But they have done it again.
Every productivity wave since the PC made work faster and piled on more of it, so companies gained more than workers did. AI does the work itself, which is why this one could land differently.
If you're using Dapper with anonymous objects to query varchar columns, you're probably sending nvarchar(4000) parameters — causing CONVERT_IMPLICIT on every row and defeating your indexes. Here's the fix.
Confessions of a naive hamster: why the smartest people in the room keep missing that intelligence is often not the main bottleneck in the real world
My thesis on the AI-first society and the future of software
Nix is a tool that takes a unique approach to package management and system configuration. Learn how to make reproducible, declarative and reliable systems.
Applying Alexis King's parse-don't-validate principle in TypeScript, where the type system fights back just enough to be annoying.
I’ve always wanted an ARM server in my homelab. But earlier, I either had to use an underpowered ARM system, or use Asahi which not only requires expensive Mac hardware but also slowed down in the past few years.
I’ve been passionate about writing “Good Code™” since before I started my career, beginning way back when I was still a kid in middle school.
A beloved hometown impersonator had no idea what he was signing up for with this photoshoot.
I do all my work in the terminal, so it has to be quick. How my zsh starts in 30 milliseconds without a framework, and how to measure where your own shell spends its time.
LedgerStore became a case study for DynamoDB, and system design publications keep praising it to this day. But the design was abandoned 3 years after it went live.
People are moving away from corporate controlled tech, and making their own. How the designs of cyberdecks are being driven by women. And how medieval guilds, Luddites, and the Arts and Crafts movement are all connected to our current moment.
Fragmented systems are brittle. Coherent systems are special-purpose. We need a new, general-purpose model so we can build coherent, multi-domain applications.
I've been testing a simple prompt with different model and harness combinations to work out which one produces best results. I do this in /goal mode.
This is probably the question I get asked most about Spacetime. It’s a simple enough question, and it certainly seems like it should have a simple answer. Scaling is a complex topic, and the devil is…
After a long day of exhausting conversations about Hybrid Post-Quantum Cryptography, random jackasses trying to play gotcha with endpoint attacks against end-to-end encrypted messaging apps, and me…
Retrospective on Project Xanadu’s success and failure: a lack of design iteration, meaningful use-cases, or practicality stopped a valuable vision from maturing into something useful. (And contrasted with my approach.)
Beginners need explicit syntax, experts want terse syntax.
The mind that builds software is far smaller than the software it builds, and almost everything good in engineering is a way of living with that gap.
101 interactive pandas exercises with solutions. Edit and run every code block directly in your browser — no installation needed.
In 2006, CHM curators traveled to Germany to follow a lead on an abandoned collection of historical computing artifacts. In a three-story warehouse, they found a real-world timeline of computing history. After days of exploring and assessing, they brought back nearly 2,000 items for the permanent collection.
I’ve recently been working on a mechatronics project that involves constructing an X-Y cartesian motion system. I opted to use stepper motors to drive each axis, and timing belts and pulleys for translating the rotational torque of the steppers into linear motion. The timing belts need to be anchored to the component…
use WASD + Mouse or tap on the screen to walk around :) Amulet of St. Michael in Kyiv, Ukraine (reconstructed with GaussianSplatting.jl, ~6M Gaussians)