For most of my career, "designer" meant I stopped at the handover. Research, concept, UI, prototype — then engineers took it from there, and a quiet translation loss took its tax on every project. That boundary has moved, and I moved with it.
The studio today
At Design Castle I now run an AI-assisted delivery loop: I design the product and drive it to working software, with AI coding agents doing the mechanical construction under my direction. Inpost.ai went from brief to the App Store this way. Nexura — a POS lifecycle platform with an admin portal, merchant onboarding and field-team dispatch — is being delivered the same way: design decisions land in the repository, not in a slide deck waiting for interpretation.
What actually changes
The obvious change is speed. The deeper change is where quality control happens. When the person who understands the user intent also reviews the implementation, the fidelity argument disappears — there is no "the dev version" of the design. Testing moves earlier too: I put working builds in front of people, not clickable facsimiles.
What doesn't change: the judgment. AI writes code the way a printing press sets type — faster than any hand, with zero opinion about whether the words are worth printing. Research, prioritization, taste, and the responsibility for what ships remain fully human jobs. Mine.
Exhibit A
This website was rebuilt exactly this way: reference analysis, a documented design language, forty projects' worth of content mined from a decade of archives, and every page built and iterated with an AI agent in the loop — under a design system I wrote first, so the machine had rules to follow. The portfolio isn't just describing the practice. It's the output of it.