93 links from the last 24 hours · 29 September 2026
An IDE for solid modelling with distance fields
Muse is getting a lot of attention — including mine — because it’s both groundbreaking technically (each user gets their own entire persistent Linux VM running in Meta’s cloud) and …
Bad news on the MX Keys Mini pickup. Usman showed up at your building around 9:15 and waited, messaged a bunch of times, and nobody came down. He left angry …
What's the highest legal FILETIME? Is it safe to use? - The Old New Thing
Being wary of how people use it and how it interacts with other types.
On Friday I gave the closing keynote at the WeAreDevelopers World Congress North America in San Jose. I tied together the key trends from the past year into a chronological …
Hey, you know it's like super obvious if you're using AI to write your scripts for TikTok and YouTube, right? [...] It's not just the general AI-isms of "it's not …
std::call_once vs. std::async - The Old New Thing
Simple vs complex machinery.
GoCode
I’m at OpenAI DevDay today, in Fort Mason, San Francisco. Same as last year I’ll be live blogging the keynote and some other notes during the day. OpenAI gave me …
macOS Golden Gate Is a Buggy Mess · SquareOrbits
In June 2008 Apple announced Mac OS X Snow Leopard, proudly boasting that it would have "no new features". Instead of piling on new things, Snow Leopard aimed to build on the success of its predecessor, Leopard, opting to focus on under-the-hood performance an…
Yes, no AI is now a feature - TDF Community Blog
One of the comments on the announcement of LibreOffice 26.8 was a simple question: “So, no AI is now a feature?” LibreOffice does not reject artificial intelligence out of hand, but an office suite used by tens of millions of people – in schools, hospitals, pu…
The long way around to screen sharing.
LLM access to models by Anthropic, including the Claude series
DevLife
Learn about shadow DOM through interactive examples that demonstrate style encapsulation, inheritance, slots, parts, and JavaScript access. Explore how shadow roots create isolated DOM trees with private stylesheets and elements …
We just shipped support for the ugliest part of HTTP: Vary
New Sonnet model from Anthropic today. They say it "runs 30%+ faster, and costs up to 30% less for most work" - it's priced the same as Sonnet 5 but …
An open source multi-tool for exploring and publishing data
To say that we were surprised at the jump and suddenness of the capabilities of our models when it came to “cyber” or “swarming” or “message boards” or anything else …
c#Code
DDD
Access large language models from the command-line
S3 Is the Future, S3 Is the Past
I'm hosting an evening event with Jesse Vincent in San Francisco on Wednesday 14th October for people who are building weird and interesting things with and on top of coding …
GoCode
Comparing exception behavior of magic statics, std::call_once, and std::async - The Old New Thing
Do you try again or give up?
Last week TypeSafe AI unveiled Jev, their first example of a new category of model that they are calling “System One models” (I’m with Maggie Appleton, I think “decision models” …
c#Code
Debugging walkthrough: Access violation on nonsense instruction, episode 3 - The Old New Thing
Trying to figure out what the interloper was trying to.
A brief history of Windows scroll bar shortcuts - The Old New Thing
Right-clicking and shift-clicking.
Datasette plugin that authenticates users against GitHub
Python web app to help rewrite your commit messages
LLM web UI for setting keys
A configurable stacking window manager for Wayland, built on River and inspired by FVWM and MWM.
C++ reminder: Function-local static variables are initialized only once, even if it looks like they get initialized multiple times - The Old New Thing
Only on first execution.
Last week, I tweeted out the three oddities of application development:
Introducing GPT-6.1 Sol
Meet GPT-6.1 Sol: near-Astra intelligence for coding, computer use, and professional work at one-fifth of Astra’s standard API input and output token prices.
dotnetc#Code
Bluesky reply bot checker
LLM plugin for accessing Jev and other TypeSafe AI models
The more time I spend working with coding agents, the more convinced I am that they make software engineering even harder. We can do amazing things with them, but unlocking …
Explain and validate SQL queries as you type them into Datasette
Magic statics vs. std::call_once - The Old New Thing
They sort of do the same thing, but differently.
Experience an interactive pixel-art celebration featuring kākāpō (New Zealand's flightless parrots) jumping and dancing to music. Click, tap, or press Space to trigger confetti bursts, balloons, streamers, and festive effects …
Test and experiment with Google's Gemini 3.8 text-to-speech API through an interactive playground where you can compose single-voice narration or multi-speaker conversations, preview the generated audio, and explore request and …
Personal
After a two year preview, Cloudflare's support for running Python code in their server-side Workers platform is now stable: "Python is now a first-class, fully supported language on the Cloudflare …
gcpCode
GoCode
It has been half a month since I started a new role at a big company. Nobody knows anything here. The specs, code, tests, PRDs, tickets, resolution of those tickets, …
The one where I’ve seen enough (for real this time).
dotnetc#Code
c#CodePolly
Every “AI” tool is missing critical features that you would need if you wanted to do real work with them. What would those look like, if they existed?
MCP was always a bad idea?
GoCode
c#Code
The Lisp family of programming languages is infamous for their parenthesis-heavy syntax. Many people, myself included, find this syntax more difficult to read than the notation of languages like Python, Java or C. The perceived unreadability of Lisp is signifi…
If you're just here for the code, you can grab it in this Gist
dotnetc#Code
No, really, you need to pass all unhandled messages to DefWindowProc, part 2 - The Old New Thing
Another example to add to the list.
Yesterday was Grok 4.7 (pelicans) and MiMo v2.6 Flash/Pro (more pelicans). Today Anthropic released Claude Opus 5.5, and around an hour later OpenAI released GPT-6 Sol and GPT-6 Luna. It’s …
When working with Win32 FILETIME, don't forget that you have std::chrono now - The Old New Thing
Let the library do the math for you.
This Next.js alternative just officially hit 1.0. Vinext 1.0 gives Next.js apps a new path to Vite and Cloudflare Workers. Let's break down what Vinext is, which Next.js features it supports, how it compares with OpenNext, and where its cache warming approach…
Accelerating the future of AI research and compute
It’s Time to Investigate the AI Labs - Cal Newport
Over the last several months, the two leading frontier AI labs have shown some brazen behavior. It started with a series of carefully planned announcements ... Read more
GitHub - firelex/jeff: Fine-tunes of Qwen3.5 and Gemma 4 for zero-shot classification
Fine-tunes of Qwen3.5 and Gemma 4 for zero-shot classification - firelex/jeff
This happened to me in 2020, and it has been on my mind again lately. During the worst of the pandemic, I found a vulnerability in a system that gave me access to the Brazilian federal system, and with that access I was able to retrieve information on any Braz…
This post is based on my talk at the Waterkant Festival 2026.
The Women Behind North Korea's IT Worker Operations - Hayden McKenzie
The Women Behind North Korea's IT Worker Operations - Hayden McKenzie
Fingerprints of Jev
Someone on Hacker News suggested fingerprinting Jev’s tokenizer. The API returns how many input tokens it billed you for, so I searched for strings that separate known tokenizers and compared their counts to Jev’s.
Pragmatic Anthropomorphism, or: How to Talk to an Autocompleting Cricket
TL;DR: Arguing over whether LLMs "truly understand" or have an inner spark misses what's actually useful when sitting in front of one. Treating an AI as a competent collaborator isn't magical thinking or naive anthropomorphism—it's pragmatic navigation of a va…
An IDE for solid modelling with distance fields
Lately I keep feeling that the internet has become a worse place. Not necessarily louder or slower, but emptier in the places where people used to be visible. Ten or fifteen years ago I could move through small personal blogs, read along in forums, disagree in…
Pirating the Pirates
Notes from the underground of film preservation.
The operating system Microsoft definitely did not make. An interactive parody of Windows 11, Microsoft 365, Copilot, updates, ads, and subscriptions.
I Have a Fetish for Rounded Corners
I've had a love affair with rounded corners for too many years. Recently, I came across my first business card, which I had printed in 2011. It had, guess what, rounded corners. I remember hunting for a printing shop that could make them that way. Failing to …
Machine's Search for Meaning
Compute allocations are the new slots of time and freedom. Does getting more of them depend on what the machines want?
Cutting LLM tokens on big spreadsheets — Andrei Kushniarou
What each step really saved, measured on a 50,000-row report, while I was building an MCP server for Google Sheets.
Editor is an extremely simple-to-use terminal text editor. It was written in C, and it offers simplicity while being very responsive and performant. Editor was …
Test-driving Jev, a decision model trained for calibrated probabilities, on identity resolution across five data sets. Compared against gpt-oss-20b, Claude Haiku 4.5, Sonnet 5 and Opus 5.5, and traditional entity resolution (Fellegi-Sunter and gradient-boosted…
I work on AI evals and reliability for agentic systems. Skeptical of tooling hype, mostly because I keep measuring it.
Predictable Swarm Scaling
Under suitable conditions, we can use simulation to roughly estimate the best structure for an agent swarm and its scaling law.
Your CI database picks different plans than production because its statistics differ. Inject production stats with PostgreSQL 18, or generate rows that carry the distribution. Here is how to tell which one you need.
What Heraldry and Japanese Mon Can Teach Us About Building Visual-Identity Generators by Ben Overmyer
Introducing Claude Sonnet 5.5
Claude Sonnet 5.5 is a clear upgrade over Claude Sonnet 5, runs 30%+ faster, and costs up to 30% less for most work.
On December 11, 2025 I had shoulder surgery. In the operating room I crossed into an alternate universe's timeline.
LWP::ConsoleLogger::Everywhere lets you debug LWP and Mojo UserAgents you don't even know about. Now we also have support for HTTP::Tiny. You don't even have to know or care which packages the UserAgent calls take place in.
Dear User of the Universe
An essay on the self-correction mechanism and its civilizational consequence.
computer art! non-24-hour sleep/wake exhibit
A high-performance and high-level purely functional data-parallel array programming language that can execute on the GPU and CPU.
10 years ago, when I started a new job, one of my first tasks was to evaluate various garbage collectors.
How I'm using Claude in fall 2026: Blender, CGI, and music composition, without running code on my computer or opening a single Claude Code thread myself.