About
the long way here
My name is Aníbal Rojas, and I am not a Software Engineer: I studied Computer Science, which looks similar but is not the same. Many years ago I used to program. I loved programming. Perl at first, then a ton of Java until I got tired of the XML push-ups that were all the rage at the time, and finally Ruby (to use Ruby on Rails), which I still consider the most beautiful programming language in the world, even though my bet these days is Elixir.
Then management "happened" to me, as it does to many developers: somebody at the company had to be the manager, and it fell to me. I did not know what being a manager was. I had no good role models, and no awareness that it was something you could actually study. In the transition from individual contributor (IC) to manager I suffered a horrendous burnout. And yet, it turned out I was very good at that manager thing.
To get past the burnout, one of the things I did was psychotherapy, and after a year of Gestalt I ended up enrolling in a formal training as a Facilitator of Gestalt Processes. When I started, I could not imagine how much growth it would bring me, or the tools it would leave me with for the rest of my professional life. Few VPs of Software Engineering can discuss the architecture of an Agent, diagnose a communication problem between teams using Virginia Satir, and redesign a process based on Eliyahu M. Goldratt’s Theory of Constraints, and I am one of those few. And since psychotherapy has no place at work, I got certified as a Systemic Coach.
I spent many years in management: again and again I led teams and helped them raise their level, take on complex challenges, each bigger than the last. I was lucky to always be involved in projects on the cutting edge: from very early on, the web itself, and later cloud computing, were part of my day to day. I still remember having to explain to a client that their team would not need to install any software: they would use the browser to access the application. I worked with big companies, oil and gas, telco, aviation, etc., and I also worked with startups, completely separate worlds.
In my last job I spent six years as Vice President of Engineering at the most important online education startup in Latin America, Platzi: a challenging, complex, demanding job, and one I will always remember because no other job ever let me give so many people a chance to grow: the people on my team, and above all our millions of students.
Since in a startup your scope has no limits, and you do whatever it takes to make the bet work, my job stretched far beyond the Engineering team and I helped build the new layers of leadership. From that experience came the Filosofía para Managers audio course, my favorite of all the ones I recorded: an evergreen course people still write to me about on LinkedIn to say thanks.
In 2025 I took a sabbatical. At 56, I prioritized my health "technical debt", and with the time left over I set out to explore, in depth, software development built on Generative AI. I had already gone all in on understanding the first principles of Generative AI since OpenAI launched ChatGPT on November 30, 2022, and from very early on I helped leadership layers at companies understand what was coming, but this time I got in up to my elbows in the mud.
Not only did I try piles of "coding assistants", of which there was a Cambrian explosion in 2025, but I also set out to challenge many of the ideas I myself held about what it meant to develop, maintain, and manage an application, and in particular that mythical task at the core of software development: programming. We now understand there is another kind of intelligence that can do it as well as we can or better, that is available 7 x 24, 365 days a year, and that it falls to us to operate and manage.
These days I am focused on exploring the possibilities, trade-offs, and limits of agent systems for software generation. For a long time I was "married" to Claude Code, then I started trying "Chinese models" with OpenCode. I confess it has been an absorbing and fun process, not free of frustration, and extremely demanding. My next challenge is exploring Pi Coding Agent, because I believe it is the simplest path to real ownership of the harness, and I see that as key to the future.
For all the brutal impact Generative AI has had on software development over the last two years, the reality is that this is just beginning. The tools will change, all of them: the models, the agents, and the harnesses that absorb me today will have other names next year, and what I am looking for is not in any of them, but in what remains when they get replaced. That is why the only way I know to keep pace is something unpopular that I insist on promoting: critical thinking from first principles. Understanding what these systems really are, what they can do, what they cannot, and deciding with your own judgment. That is what I practice, that is what I write, and that is what I bring into every room I work in.
what i actually do in a room
Strip the engagement letter down and what's left is four verbs, at different levels, with different intentions. All four point at the same destination: AI working at the core of your workforce.
Facilitate
Working sessions where a leadership team builds its own map and its own decisions. I bring the structure and the uncomfortable questions; the room does the work. The alignment holds because nobody had to buy it: they built it.
Coach
One-on-one with executives, confidential, under a contract: we define what success looks like in your words, you practice between sessions, and we close when you no longer need me. The training behind it is formal, coaching and Gestalt; the intention is the growth of the leader, not the comfort of the sessions.
Mentor
What transfers is judgment: thirty years of pattern recognition about which fires burn out on their own, and which ones take the building down. For engineering leaders and senior engineers meeting those patterns for the first time.
Train
Installing capability in teams: first-principles literacy about what these models actually are, then the two practices that survive model releases, steering and backpressure. Not tool tutorials; tutorials expire.
what i refuse
What I don't do is consulting, not in the sense the word has come to mean. No analysis that lives out its days on a shelf, no report as the product, no slide decks, no readout theater where everyone nods and nothing is different on Monday. Those artifacts produce activity, not outcomes. And outcomes require somebody changing how they work, which no document has ever done.
Part of this is conviction. Most of it is arithmetic: if what you bought is a document, then the day I leave, the change has not started. Also, I have produced enough slides for one career.
who this is not for
If you're shopping for a tool recommendation, or for silver bullet skills and recipes, this is the wrong room: any list I gave you would be stale before the invoice cleared. The same goes if you need an outside name to validate a decision already made, a deck to reassure the board, or someone to blame when the initiative stalls. I don't rent out authority. I work with leaders who intend to do the changing themselves.