Building in public, with an AI co-pilot

This site is the new home for what I do, what I'm learning, and what I ship — published in English and Spanish because the work is happening across borders.

Two things make my output today look unusual for a single operator: mobile depth and an AI-first workflow.

Mobile depth

I've been writing native Android since 2009. Since 2023, I've been moving the parts that don't need to be platform-specific into Kotlin Multiplatform + Compose Multiplatform, so I can ship one codebase, one design language, and one UX to both Android and iOS. Linki is the proof: a real bookmark manager with AI summaries, voice search, and a paid Pro tier — all running off shared logic and shared UI.

For everything else, I have the platform reflexes a senior dev develops over a decade: knowing what's a good fit for a Flow, when a LaunchedEffect will bite you, why a LazyColumn is the right primitive even when it isn't.

AI-first workflow

Since 2024 every step of how I work involves agents. Claude Code drives the day-to-day. I run my own multi-agent toolchain (research → design → code → test → ship), with Gemini for high-volume tasks and a small fleet of MCP servers for the I/O.

The honest result: throughput a small studio would envy, judgement still that of one senior. The agents don't replace taste — they remove the friction between having taste and seeing it executed.

Formalising the foundations

My applied AI master's at TheBigSchool covers the breadth around the work: LLMs and agents, MCP, the EU AI Act and AI for marketing. Using AI is already a daily reflex for me; the master gives the foundations around it.

If you're building in this space and want to compare notes, say hi.

— Erik