Hello there! my name is André, I’ve been coding for 21 years, I’m self-taught, and have been working professionally now for 19 years. I build distributed systems for a living in addition to delivering AI integration solutions and general software engineering.
Oh and I build browser games for fun and to create art, I also produce music from time to time (without AI) in addition to creating open source tools to help make coding with AI easier.
The day job is Principal Engineer working on high-throughput payment and loyalty systems in GoLang, Python and PHP with Google Cloud, I also run my own consultancy where I deliver work in all sorts of languages and frameworks from React to Swift, though I have a few preferences in each domain that I lean on… I tend to use the right tool for the job.
I use AI heavily and have stopped handwriting code almost entirely, so you could say I’m a vibe coder now :) I’ve created a second brain with all my knowledge and experience that helps me to deliver at record pace, how I use it makes all the difference. I work out the architecture before anything else, I’m careful about how state flows through the system, and rather than guessing at performance I actually measure it. The model writes plenty of the code, I’m the one deciding what’s correct.
All of my games so far run in the browser, built with Three.js, there’s no download, no launcher, you just click and play. Velocity GP is a Wipeout-style anti-gravity racer, 60fps WebGL with hand-tuned track physics and a real sense of speed, and I wrote the whole build up on Substack if you want the engineering behind it. Taken and Shellbound are two more, both made to see how far you can push a single browser tab.
Most of my open-source work is about making AI-assisted development less of a mess. Mnemos is an MCP-native memory layer so coding agents stop forgetting your codebase between sessions. Traffic Control lets you run several agents on one repo branch without them stepping on each other. And the Context Substrate Method, a structured way to feed agents the context they actually need, is past 200 stars on GitHub.
