A new preprint shows GPT-5.2 proposing a new formula for a gluon amplitude, later formally proved and verified by OpenAI and academic collaborators.
gpui: Remove blade, reimplement linux renderer with wgpu by zortax · Pull Request #46758 · zed-industries/zed
The blade graphics library is a mess and causes several issues for both Zed users as well as other 3rd party apps using GPUI. This PR removes blade and implements the linux platform using wgpu whic...
A countdown for Apple to fix the iOS keyboard or lose a customer.
MonoSketch - Unleash your ideas with ASCII
MonoSketch is a powerful ASCII sketching and diagramming app that lets you effortlessly transform your ideas into visually stunning designs.
When I saw Jensen Huang introduce the Reachy Mini at CES, I thought it was a gimmick. His keynote showed this little robot responding to human input, turning its head to look at a TODO list on the wall, sending emails, and turning drawings into architectural r…
Spotify says its best developers haven't written a line of code since December, thanks to AI | TechCrunch
Spotify credits Claude Code and its internal AI system Honk with speeding up development.
Several Ring camera owners are returning their cameras to Amazon, citing a potential breach of TOS where law enforcement is using footage without permission.
update README.md format and clarify state of the project · minio/minio@7aac2a2
MinIO is a high-performance, S3 compatible object store, open sourced under GNU AGPLv3 license. - update README.md format and clarify state of the project · minio/minio@7aac2a2
Factional Drift: We cluster into factions online
Whenever one of my articles reaches some popularity, I tend not to participate in the discussion. A few weeks back, I told a story about me, my neighbor and an UHF remote. The story took on a life of
Skip the Tips — Can You Escape the Tip Screen?
Many dark patterns stand between you and 'No Tip'. A free browser game about tipping culture, guilt, and tiny buttons.
Welcoming Discord users amidst the challenge of Age Verification
Matrix, the open protocol for secure decentralised communications
Anthropic raises $30 billion in Series G funding at $380 billion post-money valuation
Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
There’s been a mixture of hype and controversy about this new AI agent thing called Openclaw1. It bills itself as a Personal AI Assistant and, basically, you set it up with a bunch of integrations and let it loose on your life. To me it appears to be like Home…
How should we go about helping our users with their active voice?
The explainability of the rules your code introduces is a litmus test for design quality. When the explanation is long or complicated, it signals a hack, an unbounded scope, or a rejected elegant design.
In the release notes for macOS 26.3 RC, Apple stated that the window-resizing issue I demonstrated in my recent blog post had been resolved.
The cost of running a business is the maintenance of the original idea behind the actual start of the company. The value of the company is that continuous maintenance. The cost of starting the business is small, and the cost of having the idea is zero. If the …
Why the quality of your input matters more than the quality of your instructions. How single words move trajectories, why CLAUDE.md can't save a bad codebase, and the uncomfortable truth about prompting.
What to do before thinking hard
As a mathematician I spend a lot of my time thinking about stuff. To make the most out of this time, I need to be able to focus. Here is a list of questions that I have found useful to go through w…
Diffusion Models for ARC-AGI: A Retrospective
In early July 2024, I spent about two weeks applying diffusion models to the ARC-AGI abstract reasoning benchmark. It was my first ML project after taking the fastai courses on diffusion. I solved one task, nearly solved a second, and moved on before running a…
The technology industry's entire purpose is to progress the technology. Anyone who stagnates in one area will eventually work themselves out of a job.
Claude Code Doesn't Make You Better at Multitasking
Engineers are running 8 instances of Claude Code at once. That doesn't mean they are completing work 8x faster. If you must hold the context for 8 agents in your head at once, your expertise as an engineer is wasted. See https://emsh.cat/one-human-one-agent-…
Coding Agents Meet Distributed Reality
AI is about to write most of the code in the world. Most of the code in the world participates in a distributed system. And distributed code is where our worst…
How to Lie with Charts | Hands-On Data Visualization
Tell your story and show it with data, using free and easy-to-learn tools on the web. This introductory book teaches you how to design interactive charts and customized maps for your website, beginning with easy drag-and-drop tools, such as Google Sheets, Data…
A reflection on being silenced for simply being different in open-source communities.
SchwadLabs — Software That Just Works
Public data tools built to be understood at first glance. No manuals, no onboarding flows.
The Last Piece of Software
A parable about a builder, the closing of gaps, and the strange grief of getting everything you wanted.
Scaling Culture Without Dilution
How culture actually scales in distributed teams. Why systems shape behavior, goals define incentives, and bonding aligns judgment under pressure.
Last week, I wrote an article whose purpose was to discuss where I saw AI-assisted coding going. That's not what generated 200 comments worth of passionate...
Reflections on using Claude Code
This is the follow-up to my post from two weeks ago on observations using Claude Code (Opus 4.5 to start, 4.6 after it released) to rebuild...
One With The Environment
We are all products of our environments, but to what extent exactly? A friend told me he's found it more effective to change his surroundings rather than trying to will himself to change from the inside out. In the Free Energy Principle, Professor Karl Frist…
Everybody thinks they can spot AI, and when it comes to text, the favorite signal is em dash usage. I was one of those people.
Hardware usually breaks because of physics. Software breaks because of math and misunderstanding
I, like many others in the great city of San Francisco, suffer from AI anxiety. I wake up pondering the meaning of work as more jobs are automated, what skills remain important in a world where knowledge work is trivialized, how best to run a swarm of coding a…
Vercel-like development setup for Django using Caddy
Get {branch}.dev.yourdomain.com previews for Django (or any WSGI app) with a single VM, Caddy, and rsync—no Kubernetes required.
TL;DR: In this post, I present five techniques related to vertical scrolling used in <HighTable>, a React component that can display billions of rows in a table while keeping good performance and accessibility.
AI agents can be tricked into revealing credentials. SEKS fixes that with credential isolation at the architecture level.
I benchmarked 4 CLI coding agents on an NP-hard optimization problem I solved by hand 8 years ago. One of them beat me. - Charles AZAM
Claude Code, Codex, Gemini CLI, and Mistral tackle a fiber network optimization problem. Claude Code beat my 8-year-old C++ solution by 62 points.