What happens when knowledge is commoditized?
A few short thoughts on where the value accrues in a post-AI world.
Welcome to Young Money! If you’re new here, you can join the tens of thousands of subscribers receiving my essays each week by adding your email below.
Between my freshmen and sophomore years of college, I came to a disturbing realization: I was set to graduate in approximately 2.5 years. Thanks to a combination of high school credits and mandatory summer courses for football players, I was going to reach the 120 credits needed for graduation in December of my junior year. As a 19-year-old who loved every aspect of their 19-year-old college life, the idea of graduating a few months before turning 21 was my personal hell.
In my moment of panic, however, my adolescent mind conjured a wonderful solution: what if I pick up a double major?
After reviewing the requirements for two degrees, I realized that earning a second, non-business-related degree would push my graduation timeline back to its rightful place: four years.
The question, then, was what should I study for this second degree: computer science or Spanish?
My primary degree choice was obvious: as a 19-year-old dude who thought the stock market was cool and decided, from a young age, that he wanted to “make a lot of money,” I majored in finance. The choice for my other degree, however, seemed as inconsequential as it did subjective. I didn’t really need computer science or Spanish, because I had no plans to be a software engineer or a Madrid-based banking analyst (have you seen the salaries in Europe?)
Ultimately, I decided Spanish would be more interesting, because 1) most white dudes from the US can’t speak a second language, and it would be objectively cool to be able to do so at any level, and 2) it would give me an excuse to travel abroad, which I did, to Spain, and it was wonderful. Considering that three years later, half of my friends from CBS are Latin American, I would say that was the right decision.
However, for years after graduation, I still had this nagging feeling that I should have learned to code. Learning Spanish is beautiful because it allows you to engage with an entirely new group of people, but learning to code is beautiful because it allows you to engage with an entirely domain: computers.
Despite software’s tentacles being buried in every aspect of our lives, I hardly knew what an API was after graduating college. I tried, a few times, to alleviate this sense of ignorance by taking Python classes on Udemy, but I found the work too boring and tedious to gain any real traction.
Learning Spanish was frustrating, but fun, because I was talking to people, and I’m addicted to talking to people. But coding? Especially when you’re trash? It’s terrible. Staring at your terminal, that black background and white cursor taunting you with mundane, incessant blinking? It sucks. So I never learned to code, even though, in the back of my head, I always thought, “I really should learn to code.”
And then AI bailed me out.
Generative AI is bad at a lot of things. You can’t rely on it for financial analysis and due diligence (yet), because most reputable sell-side reports and and other valuable data sources are guarded by paywalls. It’s not particularly funny, except in that meta “it’s funny because it’s AI” thing. And every AI-generated LinkedIn post is laughably obvious, with their emoji-laden, hyphen-filled prose.
But generative AI is pretty good at coding.
One of the more interesting side quests I’ve been following on Twitter this week is serial-indie-entrepreneur Pieter Levels’ mission to build a flight simulator/air combat game from scratch. Pieter’s shtick is that he builds/iterates on random internet/software projects. Most are interesting, many flame out, and a few make a lot of money, and he posts the monthly recurring revenue for each project in his Twitter bio:
For the last week, Pieter’s attention has been focused on building a flight simulator from scratch. The catch? He’s using a combination of Cursor (an AI coding tool) and Grok to write all of the code. In ~three hours, he built a fully playable flight simulator just by “vibe coding,” or simply telling the code what he wanted in English, over and over again. Over the last few days, he added multiplayer capabilities for dog fights, Mars (after Elon Musk commented), and a $29.99 Stripe “buy” button for anyone who wants to purchase a F-16 fighter jet to blow up their competition. Just look at this video below, it’s nuts:
He built a functional, multiplayer airplane Minecraft over the course of a few days by “vibe coding.” Not bad!
I still don’t know how to code, but in about seven minutes yesterday, I used Claude to spin up a joking, yet functional, “game” where a stick figure attempts to (unsuccessfully) extinguish a fire by throwing money at it.
And a few weeks ago, I built (again, using Claude) an email scraper that would parse, download, and summarize 550 replies I received to an email blast I sent out to around 1,000 people, to help me more quickly identify key takeaways from the cohort. Again, I cannot emphasize this enough: I literally don’t know how to code, and I used English to build something functional with code.
I can only imagine what those who do know how to code can do with this. The speed with which someone who has a basic understanding of Python can go from “hmmm, that’s an interesting idea” to shipping a functional thing is just outrageous at this point. And these tools are only going to get better, and better, and better.
So what comes next?
Knowledge has, for years, been a moat. If you simply knew more about a particular thing than your peers, and if you had more experience, then your advantage was nearly insurmountable. But knowledge is now commoditized, and its moat is shrinking. When knowledge (or the ability to quickly acquire it) is table stakes, the only things that really matter are curiosity, creativity, and agency. Are you interested in doing the thing, can you think of the thing to do, and will you do the thing?
And this paradigm shift is just incredibly exciting. The lowering of technical barriers to entry due to AI exponentially expands the sheer number and types of people who can, with a bit of grit, build interesting products, services, and businesses. Meanwhile, taste, creativity, and intuition quickly become the best (and final) competitive advantages when technology itself is homogenized. Given the rate of improvements in LLMs over the last few years, I don’t think we’re that far away from people being able to write (near) flawless code, entirely in English. The question, then, is what’s worth building?
- Jack
I appreciate reader feedback, so if you enjoyed today’s piece, let me know with a like or comment at the bottom of this page!
“I tried, a few times, to alleviate this sense of ignorance by taking Python classes on Udemy, but I found the work too boring and tedious to gain any real traction.”
Anyone who isn’t a coder and tried to learn can relate to this.
Another trait that will gain traction in the AI-led future— authenticity.
There’s so much AI slop out there already, something’s authentic personality will help people stand out even more as AI gets better.
As a retired programmer after 30 years on the line, I see the facts you see from "the other side". When knowledge is commoditized, not only does it allow people to do and try things in a much larger domain, with more freedom, but that also means increased job competition. Until the human world improves to the point where object duplicators can duplicate whatever you have and resolve all material needs and wants, a job is a necessary mechanism for one to exchange one's ability and time for financial assets to exchange for other needed/wanted materials in life. AI also covers up mediocrity, much like diplomas have. The hiring managers have to be more competent than ever lest they hire someone who can talk a good talk but not do the good work.
There might be a day when AI can make programming not only easy but also efficient. But that will take some time. Compilers have beaten human brains in assembly programming in quality of code only in the last ten years or so, no more than 20 (RISC is easier, VLIW more difficult). Compiled C code still runs at 2x or more of the speed of python code. The raw speed of hardware can compensate somewhat, but not always. When people depend on STEM and quantitative science for a living are diminished further due to increased competition, what are people to count on to earn a living? I suspect people who can talk and master human psychology will reach even more dominant positions in society than they already are today.