Git 3.0 will make SHA-256 the new default content hashing algorithm and it will be an incomprehensibly expensive and ultimately valueless and avoidable global nightmare.
I mentioned in my Hello, world! post that I built this site using Hanami. This site may seem simple on the surface; it is, after all, just a blog. However, the backend is packed with features that help me day to day. I have built a tool that lets me cross-post…
It isn’t hard to torture an LLM. You simply extract a steering vector that corresponds with LLM “discomfort”, then artificially boost that steering vector…
I think it may be best to make all hardware and software buttons and elements slightly rounded squares and rectangles instead of circles or pill shaped stadiums, except for uncommon…
turbopuffer is pushing the frontier of search. To do that, we have to fundamentally redesign our storage architecture so the vector index is no longer primary.
AI made software cheap to build. It didn’t make it reliable. And it didn’t make it cheap to own. An old law of reliability engineering explains why good projects fail, why the usual fixes make failure more likely, and what it takes to build software that works…
A detailed walkthrough of my current setup as an indie developer for maximizing my token usage while minimizing my spend. It also covers how I keep development moving when I am away from my PC.
For six days in September, ChatGPT recommended my company to people who had never heard of it, and signups jumped 36-fold. What that week says about an economy where the buyer is an AI.
I did not start this because I wanted a screenshot for /r/unixporn. I wanted to stop moving windows around manually. That was basically the entire plan. Four days later I have Hyprland, Waybar, Rofi,
6,008 real database schemas from public repos, linted: 45.1% of foreign keys have no index, 24.5% of schemas declare no foreign key, and design tools score worst.
We are introducing Clef and Clef-flash, open-source decision models hosted on Workers AI for high-speed classification and agentic workflows. Also launching: a new reinforcement learning platform that allows developers to fine-tune decision models using their …
What if Douglas Adams was right all along? What if we are just a one big computer? If you think about it, we are just beings that try to reason based on
Today we are shipping Pi 1.0, a hardened, minimal, extensible agent harness, alongside Pi Durable, a new experimental substrate for long-running agentic applications.