Building useful systems for messy, real work.
Hi, I'm Xavier Pax (xpx). I write about AI workflows, testing, and personal tools with one standard: they should be calm, specific, and actually useful. This site is my lab for making workflows more reliable and legible.
Recent Activity
The Daily Grind
Small commits, fewer mysteries, steady progress.
AI Workflows
I build with LLMs to reduce boring work, not to make louder software. The goal is simple: fewer tabs, fewer repeats, fewer dropped details.
Sanity Testing
I treat tests as a way to protect momentum. If a system helps me move faster, it should also help me notice when I am about to break it.
Math & Logic
I like first-principles reasoning, clean edges, and explanations that survive contact with reality. Fancy abstractions only earn their keep if they stay legible.
The Personal Lab
sparkle.codes is where I publish the working notes, not just the polished result: tradeoffs, dead ends, and the small decisions that actually matter.
Selected work
Tools that make
follow-through easier.
I prototype around everyday friction: planning, review, retrieval, and the gap between a good idea and consistent execution. These projects are built to be used, not just shown.
The Open Notebook approach.
I try to document the part people usually skip: why this choice, why this tradeoff, why this failed. The writing is meant to be useful even when the experiment is unfinished.