From construction to code.

Before code, I worked in construction. In that trade, the plan is fixed but the execution changes: an unexpected problem, a missing material, a last-minute instruction. You adapt without losing sight of the result.

On a borrowed computer, I understood how an algorithm works, and what lies behind an application, not just how to use it. I solved my first katas on Codewars and was hooked right away. I found the same gesture as in construction: putting each piece in its place until the whole thing holds. A few months later, I started at HEI.

On my first applications, I had only two debugging tools: Ctrl+F and console.log. Facing a bug, I picked a word from the error message, searched for it across every file, and added console.log calls everywhere to see whether the app really went through that piece of code. It was slow. Sometimes the message I expected never showed up, and that was the answer: the code never went there. Bug after bug, I learned to follow the path an application really takes, not the one I imagined. Even today, when a bug fights back, I still reach for a console.log now and then: it's often the fastest way to see where the code really goes.

On a team project, we chose to build Klioba from the code of Tsinjo, an app that works in a similar way. We had to import that code into Poja, the deployment platform (IDP), which ran it through a series of automated checks. Every failure meant starting over, so each attempt was expensive.

One day every test passed except one: GET /ping. The logs gave us nothing. On our projects, /ping had always been public, so security was the last place we thought to look. Going back through the Spring Security configuration line by line, we finally got it: in Tsinjo's code, /ping was secured. One line of configuration, found by ruling out each hypothesis in turn. It was the console.log lesson again, one step higher: the problem was where we were sure it wasn't.

At first, I barely used AI to code. Today I use agents like Claude Code to write faster and get up to speed on a new stack: well guided, AI clearly speeds up shipping.

What keeps me going on a project, even when it gets stuck: the curiosity to try a technology I haven't mastered yet, and the fact that I don't like leaving a problem unsolved. The technical challenge itself is often what keeps me focused until the end. My tools have changed since those first console.log calls. The reflex has stayed.