Start here if you're new to ML. In nineteen problems we'll get from dot products to a working training loop, all in NumPy. Implement the building blocks (sigmoid, softmax, cross-entropy, attention masks), then build a linear layer and work through the backwards passes that make training work.

When you build software for embedded devices (your Wi-Fi router or home automation setup on your Raspberry Pi), there is always the question how to build these images and how to update them.

Some reflections on trusting trust, and how deep the rabbit hole goes.

Last year, I came agross a Linux kernel feature called TCG_TPM2_HMAC. It claims to detect or prevent active and passive interposer attackers.
It all sounds really great. We should care about interposer adversaries. It’s great to use the TPM features that were invented to help us with these problems.

Windows 3 is often said to be just an UI on top of DOS. This article presents some of the inner side of Windows 3.x and will show that it is more ambitious and advanced than that.

Pipes are ubiquitous in Unix --- but how fast can they go on Linux? In this post we'll iteratively improve a simple pipe-writing benchmark from 3.5GiB/s to 65GiB/s, guided by Linux perf.

This blog post describes an alternative, straightforward approach to collaborative text editing, without Conflict-free Replicated Data Types (CRDTs) or Operational Transformation (OT). By making text editing flexible and easy to DIY, I hope that the approach will let you create rich collaborative apps that are challenging to build on top of a black-box CRDT/OT library.

In a previous post, I’ve shown how to use the rayon framework in Rust to automatically parallelize a loop computation across multiple CPU cores.
In this post, I’ll first explain which profiling tools I used to chase optimizations, before diving into how I built a faster replacement of Rayon for my use case. In the next post, I’ll describe the other optimizations that made my code much faster. Spoiler alert: copying some data sped up my code!

Macros, iterators, patterns, error handling and match make Rust almost perfect

Fuzz testing is a technique for automatically uncovering bugs in software. The problem is that it’s a pain to set up. Read any fuzz testing tutorial, and the first task is an hour of building tools from source and chasing down dependencies upon dependencies.
I recently found that Nix eliminates a lot of the gruntwork from fuzz testing. I created a Nix configuration that kicks off a fuzz testing workflow with a single command.

Web browsers are ubiquitous, but how do they work? This book explains, building a basic but complete web browser, from networking to JavaScript, in a couple thousand lines of Python.

This guide explains everything you need to know to build stand-alone photovoltaic systems that can power almost anything you want.

The course is a hands-on, research-level introduction to the areas of computer science that have a direct relevance to journalism, and the broader project of producing an informed and engaged public $100 installment loan. We study two big ideas: the application of computation to produce journalism (such as data science for investigative reporting), and journalism about areas that involve computation (such as the analysis of credit scoring algorithms.)

In this post, we are going to take a deep dive into pointer tagging, where metadata is encoded into a word-sized pointer. Doing so allows us to keep a compact representation that can be passed around in machine registers. This is very common in implementing dynamic programming languages, but can really be used anywhere that additional runtime information is needed about a pointer. We will look at a handful of different ways these pointers can be encoded and see how the compiler can optimize them for different hardware.

In this article we build off what we’ve already learned about io_uring and extend that to build an async echo server.

The underlying problem we are facing here with all these examples is the same: having the same set of bits which can mean two different things and needing some way to distinguish those two meanings. Failure to do so leads to ambiguity at best and serious defects at worst. That's why you see so much emphasis in modern systems on type safety and on strict domain separation between different meanings.

Talk from POSETTE conference