February 2025

We’re shipping a new API in Firefox Nightly that will let you use our Firefox AI runtime to run offline machine learning tasks in your web extension.

Concluding, I think this is a very interesting way of working with AI inference in the browser. The obvious downside is that you need to convince your users to download an extension, but the obvious upside is that you possibly can save them from having to download a model they may already have downloaded and stored on their disk.

Antithesis’ ability to play like a computer, not a human being, is central both to finding bugs and beating side-scrolling shooters.

"Bookmarklets" are an underutilized and frankly little-known feature in Chrome and Firefox (and probably most other browsers too) that allows you to create a bookmark that contains a little bit of JavaScript instead of a link to a web page.

a computer club is where a group of people hang out and do computer together

If you’ve seen Crimson Tide, or the first few minutes of WarGames, you might already know what this post is about. Both films prominently make use of a one-time authentication device which the user snaps in half to reveal a card that has some secret code printed on it — and as it turns out, there are at least two different projects that aim to replicate the props used in the movies.

Build Your Own Timeline Algorithm

Timeline algorithms should be useful for people, not for companies. Their quality should not be evaluated in terms of how much more time people spend on a platform, but rather in terms of how well they serve their users’ purposes

By focusing on the answer correctness as a key success metric, and designing our datasets and metrics carefully, we’ve managed to build a reliable evaluation process which has helped us increase confidence in our system’s quality.

I've written about this a few times before, so I just started a sqlite-busy tag to collect my notes together on a single page.

The tl;dr on this, to prevent SQLITE_BUSY errors even when setting a timeout, don’t ever upgrade transactions to read-write. If you know you are going to write in a transaction, use ‘BEGIN IMMEDIATE’, or start off with the write. But do read on for why this is so, and how other databases struggle with this problem as well.

Have you ever hit the number keys while watching Youtube? That simple action jumps the video to different timestamps - and creative musicians have discovered you can use this like a primitive sampler. This observation sparked an idea. As someone who loves experimenting with drum machines like the Arturia Drumbrute, I wanted to take this Youtube "feature" and transform it into something musical. I built Youtube Sequencer using TypeScript and Redux to create a robust, reliable music-making tool that doesn't miss a beat. The result is a sequencer that turns any Youtube video into an instrument. Traditional sampling can be time-consuming, requiring you to download, trim, and import audio files. Youtube Sequencer removes these barriers - just paste a URL and start creating.

Transformer Lab is a free, open-source LLM workspace that you can run on your own computer. It is designed to go beyond what most modern open LLM applications allow. Using Transformer Lab you can easily finetune, evaluate, export and test LLMs across different inference engines and platforms.

Run your own AI cluster at home with everyday devices

The home of the Web Applets spec, demo and SDK

An AI based automatic image tagger

Mostly revealed to the Web by Solid.js, shortly after embraced by Preact, but also blended into usignal, or implemented by many others, this pattern convinced tons of developers that hooks might not be the best solution to reactivity out there, but while the concept might look and feel super simple and natural to deal with, the complexity behind spans from 30LOC to pretty convoluted code bases.

Our compact signals implementation is inspired by Andrea Giammarchi’s excellent article on signals. If you’re curious about the nitty-gritty, I highly recommend giving it a read.

Today, I learned that PyScript exists. PyScript is, at its core, some rigging to make Pyodide, a WebAssembly port of the CPython Python interpreter, real easy to use as a proper client-side web scripting language. Just include one <script> tag and now <script type="py"> is a valid thing you can have in your website. Put Python code inside of it, interact with the DOM, the works. I think that rules and/or is kind of horrifying, depending. Obviously, for me, the immediate question was: Can this run Mastodon.py? And the answer is: Yes, very competently.

The origami instructions at Origami Guide are presented in an easy to follow photo tutorial format.

thanks to the work of [decrazyo] there’s a piece of furniture that you can add to your Animal Crossing house that will never get old: an x86 emulator that boots Linux. As explained in the video below, this trick leverages the fact that Nintendo had already built a highly accurate Nintendo Entertainment System (NES) emulator into Animal Crossing on the GameCube, which could be used to run a handful of classic games from within the player’s virtual living room. But it turns out that you can get that emulator to load a user-provided ROM from the GameCube’s memory card, which opens the doors to all sorts of mischief.

Tree command, improved.

This post is about providing some guidance on how to make your simple Django app production-ready and deploying to Fly.io.

Simple and opinionated OpenID Connect server designed for self-hosters

The /llms.txt file, helping language models use your website

llm-mlx is a brand new plugin for my LLM Python Library and CLI utility which builds on top of Apple’s excellent MLX array framework library and mlx-lm package. If you’re a terminal user or Python developer with a Mac this may be the new easiest way to start exploring local Large Language Models.

To me, all signs point towards software engineering changing radically as a profession to be much more oriented around the what and why of software, and much less around the how. This will cause disruption at a massive scale in the long run. But in the short run, it's just a lot of fun to play with these tools and see what they can do.

It is surprisingly straightforward to increase the VRAM of your Mac (Apple Silicone M1/M2/M3 chips) computer and use it to load large language models. Here’s the rundown of my experiments. ... I found a way to bypass this limitation. To allocate more of your Mac’s system RAM to VRAM – in this case, up to 28 GB – the following command can be used in the terminal window: sudo sysctl iogpu.wired_limit_mb=27536

my attempt to build such a capable AI computer without spending too much. I ended up with a workstation with 48GB of VRAM that cost me around 1700 euros

link list static serverless page

You wrote an interesting comment about getting your work into the LLM training corpus: "there has never been a more vital hinge-y time to write." Do you mean that in the sense that you will be this drop in the bucket that’s steering the Shoggoth one way or the other? Or do you mean it in the sense of making sure your values and persona persist somewhere in latent space?

In a previous article, I created a concept for in-car climate controls focusing on simple, physical controls. I've now found the time to turn this concept into reality through a rotary dial with dynamic haptic feedback. Along the way, I tested what makes dials feel intuitive, explored haptic design principles, and compared my approach to what carmakers are doing today.

What if you could add PostScript code just into your HTML code and forget about all the JavasScript? What if you could configure it just with the properties of the tag, having your PostScript graphics at the size as you want and at the format you want? Like an canvas tag, but without JavaScript. Like an SVG tag, but programmable.

Still, there are setups and there are setups. Having recently visited the Swedish Amiga demo party GERP, I had the opportunity to photograph a worthy contender to - and, to me, current holder of - the throne. Without further ado, let's examine!

Django package enabling user-specific SQLite databases

Especially for side projects with limited traffic requirements and scaling concerns, the promise of SQLite is that it can remove a lot of ongoing hassles. Without a separate database, cache, or queue, there is less network traffic (because the SQLite file is local), and less servers to manage, maintain, and backup.

So the fediverse is technically open and culturally anxious about it.

Personally, I feel like I get a lot of value from AI. I think many of the people who don’t feel this way are “holding it wrong”: i.e. they’re not using language models in the most helpful ways. In this post, I’m going to list a bunch of ways I regularly use AI in my day-to-day as a staff engineer.

In this landscape, you can feel broken. But the most important truth is that your mind is not. It is a vibrant ecosystem with its own patterns, its own seasons, and its own hard-won beauty. It may never fit self-referential systems, but it was never meant to. Instead, it offers an unconventional, exciting path—one lined with subtle possibilities forming as you read this. And even if no one else can see them, remember that your world is your own. And that's enough.

A graphical music editor. A next generation tracker.

a multi-system chiptune tracker compatible with DefleMask modules

MilkyTracker is an open source, multi-platform music application for creating .MOD and .XM module files. It attempts to recreate the module replay and user experience of the popular DOS program Fasttracker II, with special playback modes available for improved Amiga ProTracker 2/3 compatibility.