Are the kids best prepared for the AI era (which is happening, whether we like it or not), the ones who understand how language actually works?

There’s so much talk about the importance of economic and AI literacy that people who never saw wise household economic decisions growing up struggle to improve their personal finances in the long term.

However, I don’t see how blurring the distinction between investment and gambling is going to help new generations improve their prospects, especially when they realize there’s little difference for them between investing in a company with real value and buying crypto or betting on prediction markets.

Platforms like Robinhood and Kalshi are not much different from sports betting; they are designed to create engagement, rewarding frequency, and bombarding users with badges and celebratory messages to make them overestimate upside over risk.

And then there’s AI

Turning financial decisions into something closer to a habit with a narrative can make inexperienced people feel that investing is like playing a video game rather than a cold-blooded, rational commitment. Something different is happening with AI: its demonization or outright banning by some adults is turning an assisting tool into what it shouldn’t be: a substitute for our most personal work, which comes from agency.

I’ve been thinking about this as I see our 13-year-old son, a bright kid with the mind of an engineer, inherited from both grandparents, imagining ways to build apps and scripts using AI; for example, a tool to automate small investments. In other words, many kids refuse to remain passive and just consume content and want to take on the driver’s seat, even when the tools they use are black boxes that conceptually resemble more of a genius sprouting from a bottle than a conceptual old-school toolbox.

Teaching teens to understand markets is a good idea. Telling them, by default, that investing in a solid company and betting on a prediction market are the same thing is not. This distinction matters, especially when regulatory guardrails are being dismantled. It shouldn’t surprise us that the platforms our kids encounter first are the ones engineered for engagement rather than judgment. What we can do is help them understand what these platforms are intending, so they’re wary of many of the perks. Nothing is for free, and many predatory behaviors begin by luring victims with candy.

A generation can be raised as informed investors or as dopamine-chasing gamblers, and right now the infrastructure is set up for the second outcome.

The scarce skill isn’t coding anymore

Anyone who has seen how fast AI tools like Claude or its competitors can analyze complex files or build websites and custom applications and scripts, got the memo: software companies and entry level jobs are going to change or vanish, and people with the ability to create richly layered prompts and a social edge might, for the first time in a long time, be in a better position professionally than more asocial, technically driven youngsters who excelled building complex, rigorous technical systems (through coding, scientific papers and experiment-driven science, etc.) in solitude.

Those who think AI is just a fad that will falter or even completely vanish haven’t seen kids prompting entire websites and apps on Claude Code and competitors, and many technical apps used by the so-called “vibe coders” are becoming the new sensations among high-school and college students. Instead of complaining on social media about the apps or games they use but don’t quite like, many of them are already thinking about their own custom applications so they can do what they envision from their computers, and perhaps soon beyond.

If young GenXers and millennials built today’s user generated content (UGC), which relied on sharing content and experiences through posts in different blends of text, pictures, and video, depending on each platform, a new type of UGC is taking over the remnants of the web 2.0 experience: new platforms will let people build their own apps, automations, and rich content through mere descriptions (and audio rants).

The lineage: Glitch, Roblox, and the UGC instinct

The concept isn’t new, though some platforms may not capitalize on the new reality for being perhaps a bit too early. Consider, for example, Glitch, a website launched in 2017 by blogging pioneer Anil Dash, Joel Spolsky and Michael Pryor, makers of the old school tools for software management Trello/Fog Creek (makers of FogBugz, a known product for old-school coders), and Stack Overflow (an issue-solving platform where users helped with responses). In a way, GenXer coders tried to lower the entry bar to the still reclusive, insider-driven world of coding, which still required people to invest time in learning the basics.

Glitch was a 2017 evolution from that old world, a browser webapp that, instead of appealing to technical users working with terminals and complex local environments, used the browser to serve a 2-screen simple, “what you see is what you get” environment that helped anyone create little apps. You typed code on the left, and the live app ran on the right; then the little thing you built was hosted for free whenever you wanted.

For the first, Glitch reduced the friction and learning curve to residual technical knowledge (which was still hard to acquire and demanded rigor), making every project not only public, but capable of “remixing”: if you liked or found useful a public app on the platform, you could click and get an editable copy of it running under a new URL; it felt a bit like a friendly community of tinkerers, a window to the “open web” the world had lost before everything got churned and packaged into 2 app stores controlled by Apple and Alphabet (which got a cut from any transaction made in their respective walled gardens).

The experiment never got the traction it could have, perhaps because it appealed first and foremost to people with a high level of technical literacy. Where Glitch faltered, another platform created to encourage people to create their own worlds, Roblox (which started much earlier, in 2004, though it took off years after), succeeded.

It’s vibe coding’s turn

Roblox shared with Glitch the UGC instinct—empowering ordinary people to build things for others—, though its ethos was not open for the web to benefit as a whole, but yet another closed garden, a walled amusement park in which the engine is proprietary, the code is hidden from the community, and the whole environment is fine-tuned to encourage monetization. And, unlike Glitch, Roblox focused on user-generated games for kids.

Still, before AI, it took a long time and quite a bit of knowledge to build things even on low-friction, community-based platforms to build apps and games like Glitch and Roblox. Then, on November 30, 2022, ChatGPT launched. And chatbots transformed the way we are going to create digital artifacts from the time being, to the point that even the most purist developers rely to some extent on the automation that one new activity has brought to the table, which pop culture has dubbed “vibe coding.”

In less than four years, the world of programming has fundamentally changed, and every time Claude (so far, the most performing LLM for technical work and companies, even if the people at Anthropic don’t quite know what Claude “is” or how it “works,” as one New Yorker piece by Gideon Lewis-Kraus) launches an improved tool to improve and speed software design, previous generation software companies offering pre-AI services get more concerned (and punished in the stock market).

Figma, for example, went from having a central role in the world of web apps and services that run the digital reality surrounding our experience online to being perceived as a product in deep trouble due to new AI offers such as Claude Design.

So much for literally “learning to code”

For three decades, many utility-centered people insisted that kids shouldn’t invest in the humanities and social sciences anymore and go not only full-on STEM, but focus on computer science or (even better for the VC-driven model) just “learn to code.” This turned out to be the most flawed advice that many people now leaving college have ever received.

Still making a big economic effort to send your kids to coding and robotics camps instead of just “camps” where they learn about the world and, among everything else, socialize with others not relying on screens? Time to choose the more affordable, just “nature and social” option. Being “grounded” has never been more strategic. (That is, unless it’s the kid who insists on going to robotics camp and is already laser-focused on it: why condition a natural interest emerging from inner agency and drive?).

And, surprise surprise: among them, those who decided to learn as much as possible about the humanities, from philosophy to rhetoric and composition, to linguistics, to anthropology and ethnography, comparative literature, history, and (uh oh) journalism. So much for the Pro Advice Industrial Complex, and a lesson to learn: interacting with interesting people, reading books, and being curious about the world (not to talk about learning how to collaborate, or build and plant things with your own hands, understanding how real-world complex systems benefit from care and strategic, low-maintenance), are skills poised to pay off in the future.

Learning to code? Not as much as it used to. Though combining the two worlds—bridging soft and hard skills, social abilities with formal rigor, humanities-driven curiosity, and engineer-driven problem solving—could become even more strategic than it already is.

The García Márquez paradox

We could explain this evolution as a paradox that has already existed in the so-called hard sciences: highly conceptual thinkers bringing breakthroughs to mathematics or physics are often very bad at physics and math (Albert Einstein was famously weak at arithmetic). Which reminds me of the struggle with basic grammar rules of many of the most celebrated writers, like Gabriel García Márquez (who said he learned to write by listening to his grandmother tell ghost stories, understanding from a young age that register does more work than plot if you get it “right”).

No wonder that people who’ve analyzed the potential impact of AI properly, like American scifi writer Ted Chiang, winner of four Nebula awards, are arguing that the new era is already revaluing the humanities. Guess what, it’s becoming clear what many of us already knew; the humanities are not a decoration or a suppletory, affected way of doing some sort of class or social signaling, but an actual CORE competency for working with the new environments, especially as AI training tools such as Mercor are paying experts of virtually any discipline to master niche expertise as well.

There’s nowhere to hide if the new tools aren’t prohibited as a whole, and perhaps the best way to prepare our children for the new social and professional environment is by reminding them that the more interesting, well-read, linguistically capable, and creative they are, the more prepared they will be to master the art of prompting. Many of them will become some sort of prompting Jedi, only we don’t know the type of prompting nor the name of the job yet, because they don’t exist, but they will.

According to Chiang—a natural lover of both storytelling and tech, perhaps the best intersectional place to be in the world that is rapidly taking shape—, the people best prepared to talk to a machine that runs on language are the people who spent their lives paying attention to how language actually works. In fact, disciplines like philosophy have been insisting on the closeness of math and language since Aristotle, if not earlier. What is old-school propositional logic, if not the foundation of LLMs.

A Franciscan sage on a Mediterranean hill

A 13th-century Catalan-speaking, Majorcan Franciscan sage called Ramon Llull was already tinkering with this very intuition, and many consider him the actual father of modern AI, from Large Language Models to generative AI. His Ars Combinatoria or Ars Magna (which reached us thanks to Gottfried Leibniz) is nothing but proto-AI.

Remember: a Franciscan tertiary of the Franciscan rule from Majorca (the same place from where another Franciscan, Father Junípero Serra, grew up, a note for the California history nerds) did more for automated ideation than any fancy contemporary coder. Of course, he was very competent in the humanities and spoke and wrote in several languages, including Latin and Arabic (as well as Catalan/Occitan, Castilian, and French). Knowing a lot about how language works is almost like intuitively knowing how complex models (math, life) work.

There are many writers who were naturally competent at building worlds from their very own registry. Jorge Luis Borges, a librarian before anything else (a Quixotic sage), liked to catalog imaginary books and understood that a description, if accomplished, can summon a world.

Joan Didion opens her story Last Words (from Let Me Tell You What I Mean, originally published in the New Yorker in 1998) with the first paragraph of Hemingway’s A Farewell to Arms, whose musicality had intrigued her since childhood:

So pervasive was the effect of this Hemingway diction that it became the voice not only of his admirers but even of those whose approach to the world was in no way grounded in romantic individualism. I recall being surprised when I was teaching George Orwell in a class at Berkeley in 1975 by how much of Hemingway could be heard in his sentences. “The hills opposite us were grey and wrinkled like the skins of elephants,” Orwell had written in Homage to Catalonia in 1938. “The hills across the valley of the Ebro were long and white,” Hemingway had written in Hills Like White Elephants in 1927.

This is the beginning Didion refers to:

In the late summer of that year we lived in a house in a village that looked across the river and the plain to the mountains. In the bed of the river there were pebbles and boulders, dry and white in the sun, and the water was clear and swiftly moving and blue in the channels. Troops went by the house and down the road and the dust they raised powdered the leaves of the trees. The trunks of the trees too were dusty and the leaves fell early that year and we saw the troops marching along the road and the dust rising and leaves, stirred by the breeze, falling and the soldiers marching and afterward the road bare and white except for the leaves.

Montaigne, the good prompter

Many writers get to such a level that they create styles, registers, and entire worlds. Ted Chiang himself, who started as a technical writer, writes as if he were creating thought experiments about what happens when you specify a premise with precision and originality. Good writing is precision behaving like magic. So, back to García Márquez, in his case, “magic realism” makes total sense in more than one way.

Oddly, society spent a decade telling kids that studying English, comparative literature, or journalism (would you even?) were worthless ways to waste one’s time. We’re certainly going to attend an interesting societal experiment in which competent people issued from the humanities will be naturally competent using advanced LLMs. Perhaps more competent than many engineers working at Anthropic, OpenAI, or Alphabet’s pioneering AI division.

The narrative arc of this section of reality works strangely like a good short story, or perhaps like a speculative essay if it were to be written by the forefather of the genre, Michel de Montaigne. By the way, Montaigne came with the denomination of “essais” because that’s what they mean, and what they are: interesting attempts to tap into the things that assemble interesting phenomena. Montaigne had exactly the “good prompting” mindset.

Remember the dismissive commentary regarding dusty, marginal college departments trying to conceive devalued humanities degrees, struggling to barely exist? Turns out their quiet devaluing can end up being very detrimental to many people naturally inclined to experiment with writing and perhaps trying philosophy, rhetoric, comparative literature, and the like. Since November 2022, when the initial, buggy ChatGPT launched, it was already clear that the scarce skill isn’t writing code; it’s specifying intent in language precise enough for a machine to act on.

One question we adults (parents, mentors, professors, tutors, competent representatives of our respective disciplines, people, citizens at last) should ask ourselves is what gets lost when an entire society spends over a decade sounding the alarm and telling its best and most disciplined students that writing or reading good books isn’t as important as Python and JavaScript. As dealing with programming gets more and more relegated to LLM tasks, the generalists are in a different perceived position now.

Three kids, three different answers

Which brings me back to my kids. My older daughter, already in college, is an avid reader and is naturally inclined to writing and other creative endeavors (perhaps she’s seen adults around her reading books, writing, editing videos all her life? What’s nature and what’s nurture?).

My younger kids and many of their friends are naturally interested in the potential of AI, and not always for the wrong reasons. Our middle daughter wanted to create an online publication so she could publish little articles and pictures on noticeable local events that she and her friends find relevant, and having assisted AI help made her confident that the task was doable, allowing her to focus more on writing articles and thinking of angles, what pictures to plan during outings, etc.

She bought the domain and hosting, changed the DNS records, and tweaked the HTML to get what she wanted. It wouldn’t have happened (not like that) with a bit more friction. To her, it felt a bit like learning an instrument by playing a song she likes, which would increase the stakes of actually making it happen (which is the fun of “learning by doing” musical programs for kids and adults: why start with solfège theory?).

As for my son, he isn’t only interested in finding potential flaws to exploit as a competitive advantage in things such as crypto trading, but sometimes likes to dream big about which conceptual tools could unleash real superpowers, according to him. Like, for example, building an entire platform (using AI, of course, though he’s already very competent in computing and programming basics, due to self-interest) for people to easily build useful applications effortlessly, by building ready-made blocks that would graphically speed up the process.

In other words, he’s imagining some sort of Glitch or Roblox, but universal and capable of allowing people to build any conceivable app or script to improve their lives. For example, sometimes he wants to create little apps for Kirsten so she doesn’t need to spend so much time on tedious work that has to do with our incredible amount of big video files, audio, etc.

Llull waving at us from the top of the hill

Such a platform to allow others to easily build useful apps, if open to all, could be democratizing, but at the same time wouldn’t empower its users (“builders,” “creators”?) as much as Glitch, which has a “learn by reading” ethos instead of a simple “describe and create” approach. There are early attempts, however, of creating this evolution of social media: YouWare, AI Studio remix, Remix.gg are some of the still-niche examples that come to mind.

We can imagine, seven hundred years before Claude, Ramon Llull, an expert in Aristotle, walking up a Mediterranean hill, ruminating on whether language and logic were the same thing. Could the right combinations of words generate some sort of “truth”? He was early, and he certainly didn’t want to fall into heresy in an era when he could be killed for it, but he wasn’t wrong on his “celestial” gut feeling.

Ramon Llull’s intellectual descendants aren’t purist coders complaining about the career intrusion of “vibe coders,” the arrivistes of our time and zeitgeist. His descendants are rather the kids reading widely in several registers and languages, paying attention to how sentences work, learning to name things precisely.

It turns out I have some of those kids at home. Some are yours. Perhaps, the best thing we can do for them isn’t paying top money for yet another coding camp. Perhaps a library card, a second language at home, a habit of argument, a long walk to foster healthy daydreaming, a stubborn belief in their own potential (the one every writer and every Franciscan and every good teacher has always held), can do more for them.

Llull would have understood.