Abstraction
Before writing a single line of code I try to see the whole problem from the exact height: its limits, its consequences, the necessary separated from the incidental. Philosophy gave me that method —analyze, define precisely, distinguish— and engineering gave me the place where the method becomes something people use every day.
Thinking well before building can be costly: it is slower at first and demands resisting the rush. But it is the only thing that holds a system up as it grows. Abstraction is not theory about practice; it is a daily practice.
How much of what we call a technical problem is, in fact, a badly stated one?
Raised by a screen
I grew up in a neighborhood in Antioquia, self-taught, with a universe that fit inside an overheating computer. The internet raised me: foreign forums at three in the morning, badly translated manuals, trial and error with no witnesses. The power, if any, was no gift; it was a strange patience —the patience of someone who talks gently to a machine until it boots.
I learned by repetition, in silence, long before I had anyone to share it with. That was the first training: boring, long, with no audience —and precisely because of that, it held.
What if depth is neither inherited nor taught, but only trained?
The long work
There was a period when I chose to step away from the noise and build in silence. Out of that focus —taken to the extreme— came the largest systems I have ever built. What entered was a technician; what came out carried a philosophical question.
I do not tell it as grief, but as method: deep work needs silence, and that silence, well used, becomes work.
What do you build when you stop building for others and start building to understand?
The word that makes you think
I learned much of what is human by moderating communities of thousands. There, rhetoric stops being ornament and becomes responsibility: there is a word that only wants to win and another that makes you think. Plato's Gorgias left me that distinction as a compass.
It is not enough to know how to persuade or execute; one must know toward what end. Mastery of the word, like mastery of the machine, multiplies force —it does not manufacture judgment.
What good is convincing if you cannot show what is true?
When an argument can be run
I wrote ST, a language where formal logic is not studied: it is executed. It has its own SAT solver, type theory and thousands of tests that hold it up. An argument stops being commentary and becomes code that runs and is verified.
That opens a question I care about more than the language itself: if formal logic can run on a machine, what does that say about thought? Proving and computing are not the same, and the border between them is the ground I work on.
Is thinking computing, or is computing only the shadow thought casts on a machine?
The agent and the judgment
I care about what almost no one on the technical side asks: the difference between the agent that executes and the chatbot that converses; between the system that decides and the one that repeats. I work with agentic AI every day and, precisely because of that, I know what it does not do.
The machine multiplies force, but it does not manufacture judgment. Judgment stays human: defining the end, holding the consequence, saying what counts as knowledge. No emergence hands that over for free.
Which part of our intelligence is exactly the part we cannot delegate?
The Thursday square
Every Thursday I open a square: a philosophical conversation among people who do not necessarily share my language. It is the humanist among those who are not, the word held in common. Warmth here is no birthright; it is a conquest, something you learn to sustain.
It is the warm counterpoint to formal logic: what remains when the formalized does not reach, when the body, the space and the other enter the equation.
Who understands a system better: the one who builds it or the one who argues it out loud?
Abstraction — the blog
Philosophy applied to the machine. Essays on how we think, how we decide and how we build intelligence. I write at the intersection, for the reader who already has both grounds.
Read the blog