I started in industrial design, which is probably why I keep wanting to build the thing.

I trained in industrial design in Moscow, where you cannot finish a project without knowing how it will be made. That habit followed me into digital work and never left: a design I can’t build is a design I haven’t finished thinking about.

Professionally I’ve spent five years in two places where the constraints were the interesting part. A six-brand e-commerce group with 25,000 products and a CMS that dictated the interaction. Then a regulated financial marketplace, where every word passes Legal and the render target still behaves like 2004.

In parallel I’ve built three products end to end — a design school, a Lisbon food guide and a film-picking tool — because there are questions you only get to answer when the whole thing is yours: what to cut, what to leave unfinished, what to be honest about.

I live in Lisbon and I’m an EU resident. English C1, Portuguese A2 and improving, Russian native.

All three of my own products were built with Claude Code, and I’d rather say that at the top than have someone find it out and wonder what else I left off.

What it did: component scaffolding, TypeScript plumbing, the repetitive parts of a build, debugging loops at two in the morning. What it didn’t do: decide that discovery should cap at three results, decide to cut a feature after eight interviews, notice that a call-to-action was below the fold, or find the XSS. Those are the decisions the work is actually made of.

At Finuslugi I use AI image tools inside the design team’s promptbooks to produce illustrations for emails and in-app banners, and an AI-assisted production workflow that took a repeatable template from around four hours to around forty minutes — with the manual QA and compliance review unchanged.

The honest summary is that AI moved my ceiling, not my judgement. It let one designer ship three products in a year. It did not make any of them good.