Hi, I'm Rafa

With the rise of AI coding tools and the whole "vibe coding" conversation, every developer seems to have a take. Here's mine.

Our job is to engineer reliable solutions -- not to write every line of code by hand. Code is the artifact, not the goal. What matters is the outcome: software that works, is maintainable, and solves real problems.

AI is the latest in a long history of layers of abstraction. We moved from assembly to high-level languages, from manual memory management to garbage collection, from bare-metal servers to containers. Each shift freed engineers to focus on higher-order concerns. The work now shifts further toward reviewing, architecting, and deciding.

The real skill is calibrating risk against velocity. An authentication flow handling tokens and secrets demands careful scrutiny. A static landing page doesn't. This portfolio, for instance -- no APIs, no auth, no sensitive data -- was ~98% written by Claude. And that's the point: knowing when to trust the tool and when to take the wheel.

AI can generate code, but it can't replace understanding. Design patterns, language caveats, memory management (even in GC'd languages), optimization algorithms -- these are the skills that let you evaluate what AI produces, catch what it misses, and know when "good enough" is the right call versus when you need to push further. But it goes beyond reviewing output. That same knowledge is what lets you ask the right questions, explain how things should be built, and define maintainable guidelines that keep a codebase consistent -- whether the code is written by a person or a model.

Software doesn't exist in a vacuum. Pushed by my lead at UMN Solutions, I've been exploring project management -- scoping work, managing risk, communicating with stakeholders. Planning to take the PMP certification next. Because shipping great software is as much about people and process as it is about code.