Programming

There's a file in your Phoenix app — thousands of lines, and the bigger it got, the faster it grew. Upfront modeling was never the fix. Your domain has rivers: real boundaries you can't see on day one. This is how you find them, and split along them without a big-bang PR.

A lot of effective engineering comes from the accumulation of small nuggets of knowledge.

Tclers wiki

Switching away from radians makes code simpler, faster, and more precise.

We rebuilt the system that tracks every dollar and bitcoin at River and swapped it in live with zero downtime.

One of the best things about LLMs is that they'll write code for you, all day long. Who cares about DRY? You don't have to be the one updating the same long con...

CUDA does not need to be installed as one large system toolkit. In this post we build a terminal Mandelbrot renderer with CUDA packages from conda-forge, Pixi, and pixi-build.

Look at what your database actually spends its day doing. In most applications, a huge share of the query load is reads of data that barely...

The Phoenix UI with a real API. Accessible, unstyled HEEx components you command from the server or the client, powered by Zag.js.

Concurrency gives us more workers; demand tells us when to stop sending them work. This is what I want to point people to when we talk about "demand".

I have spent a large portion of my career working in Java. In that time, you get used to huge classes. New functionality? Just add a new method and field to the class. The cost of each new field is rarely considered. Performance is often considered from a classic computer science perspective by considering asymptotic analysis of the algorithms and data structures in-use.

Profiling is an essential part of any software project where performance is a requirement. Computers are complex things and we layer abstractions and indirections on top of them that make things hard to reason about. At the end of the day, hard numbers are needed to discover where the issues really are. Luckily, there is a lot of great tooling for both benchmarking and profiling. Some languages (like Golang and Elixir) have profiling tools built in, and at this point pretty much all ecosystems have libraries that will provide what you need. These tools are generally pretty easy to use, but they also generally assume that you don’t leave the language.

Building the classic distributed rate limiter, using the HRW library instead of the traditional ExHashRing.

An abandoned dependency was blocking a major upgrade I needed. Here are the two ways I retire one, and when a shim beats a big-bang rewrite.

As I experiment more with AI code generation, I've hardened LocalCents with more automated Elixir quality checks than I've ever run. Here's the whole guardrail setup, each linked to how it's wired in the real repo.

Developing CLIs and TUIs in Go is fun, and there are really good packages out there to make it so. And when it comes to terminal apps, there is an amazing library called Bubble Tea to build beautiful

Coding the 1993 Paper 'The Lattice Sieve' To Learn Algebraic Number Theory in C

Mishka Chelekom v0.0.9 makes Phoenix UI AI-native - a built-in MCP server that lets AI agents generate and use components correctly, a headless component layer ported from Base UI with full WAI-ARIA, the new Kit DSL to reuse and restyle any component, comprehensive LLM usage rules, and new dock, shape, and stat components for Phoenix and LiveView.

DurableServer provides durable, distributed GenServer processes backed by object storage.

The Annotated Triple Product Matrix Multiplication Algorithm for Fourier Transform-like Matrix Products in Python

The Mojo programming language

Property Based Testing and fuzzing are a deep and science-intensive topic. There are enough advanced techniques there for a couple of PhDs, a PBT daemon, and a client-server architecture. But I have this weird parlor-trick PBT library, implementable in a couple of hundred lines of code in one sitting.

Tclers wiki

Spotify 1 million tracks dataset

This is part one of a series where I document how I’m making this site more interoperable with IndieWeb tools and standards. I’ll update this notice to link to part 2 once I’ve written it Interoperability is the heart of what makes the digital world work. In my first job I worked with academic libraries to make research more openly available, sharing data about all forms of research outputs across the globe through the use of open data formats. These community-owned standards ensured that thousands of different systems globally could communicate, making each other aware of new findings in a field, and bringing the gobal academic community into closer contact.1

How Knives Out teaches Bayesian thinking (without you realizing it)

Every machine connected to the Internet has an address called an IP address. Originally, these addresses were 32-bit integers (IPv4), giving a theoretical maximum of about four billion distinct addresses. We are all familiar with these addresses (e.g., 192.168.0.0). There was a big fuss about how we would run out of addresses. It never happened … Continue reading Parsing IPv6 Addresses Crazily Fast with AVX-512

A description of HRW/rendezvous hashing and the HRW elixir library.

A best-practice Elixir library for interacting with the bunny.net API