2025 Recap - The Year I Didn't Give a Damn About People

Translated from the Spanish original with AI assistance. The original was written entirely by hand, no AI involved.

Even though I never considered myself a great programmer, I always loved to code. And while my professional “superpower” has actually been helping the people I work with in software development grow… I CONFESS THAT THIS SABBATICAL YEAR I ENJOYED THE HELL OUT OF NOT HAVING TO DEAL WITH A SINGLE PIECE OF HUMAN DRAMA — it was glorious.

This probably sounds horrible.

But spending hours exploring Cursor, Windsurf, Lovable, Bolt, etc., and finally Claude Code, and having none of these agents/agent systems get sick, go through a family crisis, ask for a raise, or resign… was magical. Every now and then a service degradation or an outage, because the reality is the industry is still learning to handle this new kind of computational load at scale — but that was it.

And going from being a Senior VP of Engineering with dozens of people under my responsibility to something like:

Hiring → $ touch .claude/agents/hire-new-agent.md
Firing → $ rm .claude/agents/agent-that-doesnt-fit-anymore.md
Vacation → $ mv .claude/agents/agent-not-needed-now.md .claude/parking
And so on…

Where the only real “drama” was the inference limits tied to Anthropic’s MAX plan — it was refreshing.

But getting stuck at the extremes is neither healthy nor useful.

Because obviously this “nirvana of detachment”, this enjoyment of freedom from the ties and burdens that come with managing teams, was a swing to the other extreme after six intense years as Senior VP of Engineering at a startup. It’s not really congruent with my identity, my knowledge, or my skills, because not only do I deeply enjoy authentic, meaningful contact with people — it also nourishes me and satisfies me enormously as a professional.

Watching the people I dedicate my time and energy to grow in ways they — we — never would have imagined is the greatest professional satisfaction I have harvested over my 30+ year career. If you, reading this, are part of that group of people: thank you for your commitment and for taking the risk of growing.

I didn’t become a hermit either.

I had a great time alongside other people exploring the limits of the models and marveling at what we managed to get them to do. I had many conversations with company, product, and engineering leaders who came to me for advice for this very particular moment in which Generative AI is radically redefining our relationship with technology.

Honestly, having had the time — precisely during this dizzying year — to do a deep exploration of what software development means in the context of Generative AI was a privilege.

But I built up a gigantic amount of frustration.

Discovering that an enormous number of people with vast experience across the different dimensions of software development don’t have the faintest idea why a coding assistant starts producing bad results, and they resolve it by anthropomorphizing it: “it went dumb”, “it got stupid”, “it’s lost”. AND THAT SHIT WILL NOT STAND — I refuse.

That’s why I committed to explaining coding assistants from first principles, because this is only beginning, and the recipe that works today will be obsolete in three weeks. But if we approach these tools from first principles, it doesn’t matter how the models change, or the tooling, or the system prompts, or the memory mechanisms, etc… We will always be able to integrate those changes into the new workflows we have to develop.

I could also confirm that even at the end of 2025 there are companies that literally ban the use of coding assistants without a good reason. That software development teams are split between the people who will move heaven and earth to get an LLM to crank out the code — even if at first it takes them ten times more effort than doing it with their own hands — and those who believe Generative AI is shiny beads and little mirrors. That there are leadership teams demanding explanations for why, if we’re paying 20 USD for this or that tool per dev, we aren’t shipping 10x faster and with fewer bugs — and managers who don’t know what to do, or how to explain any of it.

By the way, my sabbatical is coming to an end, and starting January 2026 I will be opening a small cohort for mentoring and/or coaching (they are different processes) for people in the leadership layers — not tech-specific. If you’re interested, drop me a message here so I can consider your participation.

I will also be opening a calendar for consulting, this one specific to managing technology teams and/or AI adoption in the organization. Likewise, drop me a message to start the conversation.