My 2025 Ai Retrospective

2025 was by far the most interesting year of my tech career. Everything moved so fast in the industry, which convinces me that we are experiencing a once-in-a-lifetime shift in how we build, manage, and interact with computer software.

The emergence of Agentic coding

By the end of 2024, while working for Optibus, all we talked about was how handy the autocomplete provided by Copilot was. I quickly developed the habit of writing a clear intention of a function using comments and waiting for it to suggest an implementation for me. It felt like magic!

By February I started to use Cursor, and that allowed me to stop using the “comments hack” and actually type in the chat bar what I wanted it to build. Back then, it was only working in the scope of the open file or its tests.

Write a function that will read the format of the Authentication Headers of the received request and return which authentication mechanism should be used.

It still felt like I was piloting a machine. I was still telling it exactly what to do.

I believe it was around May, when my employer at the time was paying for a Cursor and a ChatGPT license, that I tried Claude Code for the first time. I was intrigued by its CLI focus and decided to pay for it from my own pocket to give it a try.

A couple of weeks later, I was restarting old forgotten projects, executing big refactorings in some, and putting into practice some ideas that I never had the courage to start.

The transition from coder to Architect

The more I used Claude Code, the more I trusted it, and the more I was giving it space to figure out on its own how to achieve what I wanted. Hallucinations were common, but despite the frequent course corrections, I still felt that I was more productive than before.

As I became a better AI user, I started to spend less time typing lines of code and more time planning and describing solutions for high-level problems.

Fast forwarding some months, I joined Mighty Networks, a company that had fully adopted AI Augmented Development practices. By then I had already my own set of MCPs, a curated CLAUDE.md, and a pretty good understanding of how to make good use of the tool. Still, joining an environment where there was full support and incentive to explore AI at its fullest was fundamental for my complete conversion to an AI convert.

The full support in this context meant:

• My employer sponsoring enough tokens, which allowed me to use multiple development sessions in parallel
• An internal community sharing experiences and best practices
• An optimized development workflow with AI solutions integrated in different steps of the process (development, code-review, troubleshooting…)

Now, when I look back at the last months, I can confidently say that I am way more productive than I have ever been and that I have never written so many lines of code, as paradoxical as it seems.

The question to be answered in 2026

Will I miss writing code by myself?

The answer is a topic for another post. :)