I'm Sanath, a software engineer from Sri Lanka, based in Berlin. Twelve years shipping backend systems, the last two of which have changed how I think about what "productive" actually means.
I use AI in every part of my engineering work. Not as a shortcut, but as a pair programmer I direct the way a senior engineer should. These days I'm also building agentic systems: how memory, context, and long-horizon reasoning actually work when you move beyond demos into production.
how I use AI
I plan before I prompt: what I'm building, the constraints, the patterns to follow. Without that, the AI generates fast and wrong.
I treat it like onboarding someone into the codebase. Unclear boundaries, inconsistent patterns. The AI struggles with that the same way anyone would walking in cold. So I keep the code in a state where I could hand it off. That someone is often the AI.
My job is to hold the wider picture: how it fits the system, what the tradeoffs are, when the output is subtly wrong. That's where the twelve years show up: not in typing faster, but in knowing when to push back.
Longer version: AI Won't Replace You — But a Developer Using AI Will →