Curious builder, pragmatic by choice. Every project here is a small experiment — staying close to the part I actually love: building things, and understanding how they work.
Four systems built because the off-the-shelf version was too slow, too rigid, or didn't exist. Not demos — still running.
A study app I built to teach myself. Exam simulations, practice mode and flashcards, installable and fully offline. It imports Directrev-format .md datasets for free. Private datasets (MuleSoft Developer 1, AWS DVA-C02) are available on request. Made to learn, kept because it works.
see moreThe repo lived on a network share and Git over it was slow, so the computation moved to where the data already was. One working copy, many stations, near-local git status from anywhere.
running, not publicUnifies server logs with the diagnostics a client agent collects on users' machines, in one timeline. Ships an MCP endpoint so a coding agent can read both live while it iterates — the thing that makes AI-assisted debugging verifiable instead of vibes.
running, not publicOAuth2 with proactive refresh behind a swappable interface, OpenAPI REST surface, a C# desktop client. In production it ships a 2 GB+ client update in about two and a half minutes.
running, not publicClick through — they're live, not screenshots.
A cooperative mushroom hunt in Calabria's Aspromonte forest. Learn your species from Luigi, cook for Memmo, dodge the wild boar — and the forest ranger. Three services sharing one set of rooms and accounts, live.
play itBuild Fast, Kill Faster. An Age-of-Empires-shaped skirmish that runs entirely in the browser — Three.js for the battlefield, bots for when nobody's around to raid.
play itPaper trading that behaves like the real thing. I built it to learn trading by doing: operations modelled on live markets, no real money at risk. As a pure game it's weak, and that's the point. It's a learning tool, which is also why AI agents can plug in and coach a round.
try itTech should adapt to people. Not the other way round.
"Nobody wakes up wanting to learn a new interface."
Model the system on the human who lives with it: their habits, their rhythm, what they actually care about. Not a cold, generic template. Apps, IoT devices, smart homes, business processes: the rule is the same everywhere, they're all part of the interface. Be quiet when they're busy, be there when they're not, and never ask them to read the manual first. Good technology augments a life in natural, organic ways. Think houseplant, not dashboard: it grows toward you.