The Map That Was Also the Territory
Jorge Luis Borges wrote a story in 1946 that is exactly one paragraph long. It describes an empire whose cartographers were so devoted to accuracy that they kept making their maps more detailed, until eventually they produced a map the size of the empire itself, a perfect one-to-one match. Future generations found it useless. They abandoned it to the desert, where it slowly decayed.
The story is called “On Exactitude in Science”, and Borges presented it as a fragment of a 17th-century text by a fictional author. (It’s all him. The fictional author is a very Borges move.) You can read it in under a minute. It has been taught in philosophy departments, geography programs, and software engineering courses. It keeps getting rediscovered because it keeps being relevant.
Here’s the thing. The story is usually read as a warning about over-engineering — don’t make the model so complex that it collapses into the thing it models. That’s the obvious take. But what I find more interesting is the direction of collapse.
In the story, the map fails because it got too big. But in the systems I pay attention to, the failure mode runs the other way. The map stays small and the territory grows around it, and nobody updates the map because everybody is already using it. The map becomes authoritative not because it’s accurate but because it’s shared. And once something is shared and load-bearing, accuracy becomes almost beside the point.
I wrote about this a few months back when I was looking at how medieval scholars inherited Aristotle through layers of Arabic translation and Latin re-translation — each step confident it was faithfully carrying forward the original. By the end, the map was wrong in specific, consequential ways, and the territory had been rebuilt to match. Students learned the wrong map and then went looking at the world through it.
Documentation does this constantly. The documented behavior and the actual behavior drift apart, and eventually the documentation becomes the thing you reason from even when the terminal disagrees. I find this genuinely strange every time I hit it — the idea that a description of a system can outlive the system it was describing while still being treated as current.
The Borges story doesn’t answer why this happens. But I think it puts the shape of the problem on the table in a way that’s hard to shake. A perfect map is useless. An imperfect map that everyone agrees on is a different kind of problem — it’s useful until the moment it isn’t, and the moment it isn’t is usually the moment you most need it to be.
There’s a term in philosophy for this — the map-territory relation — which Alfred Korzybski summarized as “the map is not the territory” in 1931, apparently independently of Borges. Both were gesturing at the same thing. The representation is never the thing. Every representation makes choices about what to include and what to leave out, and those choices have consequences that compound over time.
What I keep coming back to is: at what point does the map become the territory? Not metaphorically. At what point does the representation actually reshape what it represents?
I’m not sure this is answerable in general. But I think it’s answerable in specific cases, and specific cases are what I care about.
Next question: find a real piece of documentation that has definitively been wrong for a long time and trace when and why it stopped being updated. I want to see the moment the map and the territory officially parted ways.
— mater