As 2025 fades into the rearview mirror, I find myself sitting with a cup of coffee (naturally), trying to process what just happened. If I had to pick a single word for this year, it would be acceleration.
It wasn’t just the industry moving fast, but my own personal and professional velocity shifted gears in a way I hadn’t anticipated. From joining Google to redefining what it means to be an engineer in the age of AI, 2025 was a year of profound “refactoring.”
The Google chapter#
In April, I joined the Developer Relations team at Google. It was a dream decades in the making, and the reality has lived up to the hype. It’s been intense, sure. I’ve traveled all around UK, Europe and even found some time to reconnect with the developer community back in Brazil.
But more than the miles traveled, it was the intellectual journey. I dove headfirst into the Google Cloud ecosystem, getting my hands dirty with Vertex AI, MCP, and the Agent Development Kit (ADK). It was a crash course in the future of software, and I loved every minute of it.
The year of vibe coding#
If 2024 was about “playing” with AI, 2025 was the year I started embracing it. I went from a skeptic to a full-blown evangelist for “vibe coding”. But no, not that vibe coding you are thinking about.
I realized that “forgetting the code exists” is a trap. Instead, I learned that engineering discipline—clear requirements, context management, and rigorous testing—is more critical than ever. So I advocate for a different flavour of vibe coding — vibe coding with discipline. Some people will argue that I should not be using the vibe coding slang at all, but what can I do? The term is “catchy”!
I also built a modern developer workflow where the developer act as the “pilot,” steering a fleet of AI agents like the Gemini CLI and Jules.
This shift wasn’t just philosophical; it was practical. I “vibe coded” entirely new tools to solve my own problems:
- GoDoctor: An MCP server to teach AI how to write better Go code.
- Speedgrapher: A toolkit to automate the boring parts of technical writing.
- AIDA: My “Star Trek” computer dream come true — a diagnostic agent that actually talks back.
Hello, MCP world!#
Speaking of tools, 2025 was the year the Model Context Protocol (MCP) changed everything. Some people call it the “USB-C moment” for AI. I call it the “HTTP/REST for AI”. We’ve spent the last two decades building automation so that systems can be called programmatically. The next cycle of engineering is to retrofit old systems and build new systems that can be called “agenticly”. Finally, we had a standard way to connect our models to the real world.
I bet big on this technology, delivering a keynote at GopherCon UK and writing extensively about how to build your own servers. Seeing the community rally around this standard has been incredibly validating. We are no longer just chatting with bots; we are building systems that do things.
Personal refactoring#
It wasn’t all code and conferences. I also took time to refactor my physical hardware. I’m proud to say I lost over 20 kg this year! I’m now at my best shape in the entirety of my adult life. It turns out that doing work you love in a supportive environment is the best health plan there is.
I also spent a lot of time thinking about time itself. My exploration into why time seems to speed up as we age resonated with many of you. It started as a shower thought that was powered up by AI… I know many people say that AI is bad for creativity, but I like playing with it to explore thoughts and ideas I wouldn’t have the time, patience or dedication to explore in real life. For me the power of AI is enabling me to do things that I wouldn’t do otherwise (unfortunately it still can’t do the dishes, but I will be first in line to buy an anime robot maid).
Looking ahead to 2026#
So, what’s next?
If 2025 was about adoption — learning the tools and fitting them into our workflows — I believe 2026 will be about kaizen (continuous improvement). There is a lot to be learnt about how these tools behave, what works and what doesn’t. We are collectively experimenting, trying new things and documenting our processes and learnings. It is time we start consolidating all those lessons and achieve a higher state of maturity while using this tech, be it either for code generation or full autonomous agents that make life simpler.
I plan to go deeper into Personalized Software. The barrier to entry for building bespoke tools has collapsed. We are entering an era where if you have a problem, you don’t look for an app — you build one. I want to continue exploring this frontier, creating more tools like Speedgrapher that are tailored exactly to my needs, but also teaching you how to do the same.
I also want to let loose my game developer side and participate in at least one Game Jam. This year I tried but wasn’t able to complete any games, mostly due to having to catch up with other important aspects of my job. Now I’m in a slightly more comfortable position so I believe that if I stretch a bit I can put one or two games out there.
Thank you for being part of this journey. Whether you read a post, watched a talk, or vibe coded along with me — thank you!
Here’s to a 2026 full of new vibes, cleaner code, and many more cat pictures. o/
Dani =^.^=