I'm Alex.
I've spent 10 years building software for the web. For the last four, I've been obsessed with one question: how do you build AI systems that work for real, not just in a demo?
Right now I'm helping drive the shift to AI inside my company. I explore tools, design patterns, build harnesses, and help other teams separate what is useful from what merely looks impressive.
I work inside the real codebase, on what actually ships. Not from the outside with a slide deck.
I design the architecture and the controls: what the agent does, how far it goes, and how you verify that what comes back is useful. The question I'm working on is a concrete one: how much can be delegated, and what has to exist so it doesn't become expensive later?
My short answer is that an agent reviewing its own work is not a guarantee. The pressure has to come from outside. I publish the rest here, including the code and the failures.
Part of that work is akrcontext, open source, with a client running it in production.
Before code, there was design. From an early age, if something had shape, colour, or composition, it caught me. That obsession with the visual led me to the web, where the technical and the visual live together.
And before software, I spent 16 years teaching people to dive. People with no experience, in an environment where none of your surface instincts help and where mistakes don't get corrected on the fly.
When I'm not in front of a screen, I'm in the mountains. I run trails. Same logic: preparation, progression, no shortcuts.
Aleksandar Labs is my public lab. I don't publish to recap what happened this week. I publish when I've built something, broken it, measured it, or been forced to change my mind.
If you're sitting with that question, let's talk.