Undecided architecture is technical debt
Every consequential choice becomes an Architecture Decision Record: context, options, decision, consequence. When someone asks “why is it like this” in year three, the answer exists.
We’re a small team that builds web apps, mobile apps, and desktop software — and websites when they’re needed. We stay deliberately small so the people who talk to you are the people who build your product, and the person who answers your first email is the person accountable for your last release.
Levarlux started in 2022 as a web designer making sites for clients. In 2024 it turned into a development studio — mostly because my brother could code (and edit), and we kept getting asked to build the things we were designing.
Since then two more developers joined: one from around here, one from the Philippines. We’re still deliberately small, so the person who answers your first email is the same one accountable for your last release — and these days most of our time goes into apps, not websites.
Not values on a wall — constraints that show up in how we estimate, review, and hand over.
Every consequential choice becomes an Architecture Decision Record: context, options, decision, consequence. When someone asks “why is it like this” in year three, the answer exists.
LCP ≤ 1.0s, p95 < 50ms, 0 CLS, 42MB idle. Numbers get enforced in CI. If a number was never agreed, the project was never specified.
We are too small to hide people. The engineer who scopes the work writes it, reviews it, and carries the pager for it — which makes everyone write fewer excuses and better code.
Reproducible environments, documented pipelines, no proprietary build magic. Lock-in is a business model; we do not want one.
The same care applies to a status bar on iOS and a migration script on Postgres. Platform conventions are respected, not flattened into one mediocre cross-platform compromise.
If a deadline is at risk, you hear it the week we know. A studio that only reports green is a studio that has stopped measuring.
SaaS products, dashboards, and internal tools that stay quick even when the data gets big.
iOS and Android apps that work offline and feel right at home on each platform.
Mac and Windows software for work that needs real speed and access to your machine.
Marketing sites and landing pages — the side quest. Built properly, just not first.
“They spent the first two weeks arguing with our assumptions — politely, and with diagrams. That argument saved us a quarter. The system they handed over is the first one our own team has been able to extend without calling anyone.”
No account managers, no calls with someone who has never shipped anything. Thirty minutes with an engineer who builds for a living — usually all it takes to know if we’re a fit.