Writing an LLM from scratch, part 13 – attention heads are dumb
A pause to take stock: realising that attention heads are simpler than I thought explained why we do the calculations we do.
The latest blog posts to take off on Hacker News. 21,819 posts.
A pause to take stock: realising that attention heads are simpler than I thought explained why we do the calculations we do.
See how we slashed PyPI’s test suite runtime from 163 to 30 seconds. The techniques we share can help you dramatically improve your own project’s testing performance without sacrificing coverage.
I think I figured out what’s stressing me out about programming in dynamically typed languages: It’s about always being unsure if using a library in a particular way is going to work and if minor version upgrades aren’t going to break your code.
Choosing fairly when you don't know how many you're choosing from.
Why splitting your codebase too early can quietly destroy your team’s velocity — and what to do instead.
Solid Queue is Rails’ default background job backend: jobs live in your database, no Redis. See how jobs, workers, and the supervisor work under the hood.
Explore how to generate and edit images with Gemini 2.0 Flash, now in preview in Google AI Studio
Postgres 18 introduces Asynchronous I/O (AIO) that can dramatically improve read performance, especially in the cloud. Learn how these changes and the new io_method setting work and see why our benchmark results show that io_uring is the recommended setting for maximizing I/O performance in Postgres 18 over the…
From the Zed Blog: Zed is now the world's fastest AI code editor. What's more, Zed's new AI features are also open-source, just like the whole editor.
Building a fully functional, code-editing agent in less than 400 lines.
Looking at C++ from another angle can create new possibilities using Rust.
Explore the new Gemini 2.5 Pro I/O Edition, featuring enhanced coding performance, video to code capabilities, and improvements for front-end web development.
Take a look at the following code.
Mastering the full stack isn’t just a technical feat: it’s a human challenge, and no, you can’t git clone your way out of it.
Because manually squinting at hex dumps is so last century. Let me show you how AI-assisted debugging is leaving WinDBG's command line in the dust.
Go applications can implement graceful shutdown by handling termination signals (SIGTERM, SIGINT) via os/signal or signal.NotifyContext. Shutdown must complete within a specified timeout (e.g., Kubernetes’ terminationGracePeriodSeconds)…
meshoptimizer implements several geometry compression algorithms that are designed to take advantage of redundancies common in mesh data and decompress quickly - targeting many gigabytes per second in decoding throughput. One of them, index decoder, has seen a significant and unexpected variance in performance across…
Exploring Lisp's fine-grained type hierarchy.
Oberon, the jewel by Niklaus Wirth and Jürg Gutknecht: Operating System, Compiler and Computer.
This is the fifth post in my planned multipart series on memory management. You will probably want to go back and read Part I , which covers C, parts II and III , which cover C++, and part IV , which introduces Rust memory management. In part IV, we got through the basics of Rust memory management up through smart…
A look at how and why we migrated from Next.js to Ruby on Rails.
A deeper dive on our findings, what went wrong, and future changes we’re making.
Title says it all, I seriously dislike jq 's convoluted, impossible-to-remember ad hoc DSL that instantly joined heaps of misery like CMake and gnuplot in my heart. It should really have been a library with a CLI wrapper around eval in a well-known scripting language like Python.
How Jujutsu's flexibility and safety changed my approach to version control.
One year ago, I started programming a puzzle game from scratch in Zig. While it’s too early to announce the game (more on this later in the year), I want to share some of the insights I’ve gained about gamedev in Zig so far.