LLMs are the Excel of words

It’s been so long that I think I can start with “Once upon a time.”

Once upon a time, there was a middle school teacher who kept a desktop computer in his lab, in the school basement. The computer sat in the center of the classroom, a little past the first row of desks. The computer was also the focal point of the classroom: the black screen with green text—if I remember the color correctly—the sheer size of the machine, and the space it occupied all underscored its importance.

On that screen, Professor Costa showed us things: things created using the WordStar text editor, things within the operating system, and things—the most interesting ones for me—within a Lotus 1-2-3 spreadsheet.

It wasn’t long before the first real computer arrived in our home: it was an Olivetti. We’d had a Philips MSX a while back, but the timing wasn’t right: I was too young, and the MSX was too basic for a child.

But even before that, there were always various calculators. I imagine just about everyone had a calculator at home. We, in fact, had more than one: I remember a very minimalist one, with basic functions, made of matte plastic and so remarkably simple that Dieter Rams could have designed it. Then I remember that my dad gave a scientific calculator to my older cousin, the first in the family to attend college.

I don’t know when my dad bought his first calculator, but the importance and usefulness of that tool were clear. I remember that its presence was, in a sense, taken for granted: that’s what happens when a new technology disrupts the status quo, pushes aside a whole body of knowledge, and demands and introduces new ones. My dad’s generation, born in 1949, was the first to have access to affordable electronic calculators, within reach of most people: doing math wasn’t a problem anymore.

I remember being really impressed when Costa showed us how to use Lotus: not only were calculations no longer a problem, but redoing the same calculations when the initial values changed no longer took any time at all. That’s it—this was truly, truly amazing: Lotus, so to speak, shortened the timeline. Not in general, of course—the autumns and springs of old remained outside of it—but when it came to doing calculations, well, something had happened there: once done, the time needed to redo them was no longer required; it was gone.

It must have been because of that sense of wonder—and then, once computers really started coming into our homes as tools for studying, working, and entertainment, I used Excel for just about everything.

I remember studying it thoroughly. Not that I knew all the formulas by heart, but, well, I was always discovering something new. I remember that once, early in college, when spreadsheets had become capable of creating graphs, I made one to plot the Joukowsky transformation—which is a remarkable concept that explains why airplanes stay in the sky. And even in that case, time took a different turn: I’d change the parameters, and Excel would generate a new graph, a new drawing. Not only had the time needed to redo the calculations disappeared, but so had the time needed to redraw the graphs. Imagine that.

I’ve often reflected on these things in 2025, which was the first year I actually used LLMs almost every day. I’ve often found myself looking at what’s happening in the field of artificial intelligence with a sense of wonder similar to the one I felt when I first used Excel.

I’ve often thought that LLMs are the Excel of words.

I’m not talking about form: there’s nothing in the user experience that links LLMs and Excel (in fact, LLMs are still pretty clueless when it comes to calculations). And I’m not just referring to words in Italian or English: I’m especially referring to the vocabulary of programming languages, which LLMs are good at using.

I am referring, in fact, to the compression of time: before, it was only us—human animals and a few rare non-human animals—who used time to string words together.

Now we’re no longer the only ones who know how to do it. And the timeline has shrunk again.