Welcome to Young Money! New to my blog? Join 60,000+ subscribers by adding your email address below:
Blogging again after a two month hiatus, since apparently there are more listeners than readers in the world these days! Today’s essay is a ~20 minute read and a ~25 minute listen. Catch the audio on Spotify or Apple Podcasts.
Two months post-book launch, it’s time to start blogging again. Does anyone blog these days? Does anyone read blogs these days? I don’t know. The “death of reading” is certainly a popular hypothesis in 2026, with fried attention spans, TikTok, screen addiction, and the like, but that’s a topic for another blog post. I still find writing to be the best way to distill jumbled thoughts into concise ideas, so it’s time to resume that practice.
This is a blog post about “AI.”
Of course, every blog post written by someone at least adjacent to the tech sector is about “AI,” so that’s not particularly interesting or novel. So this isn’t really a blog post about “AI.”
This is a blog post about indecision and impulsiveness and adaptation and the fact that, perhaps, friction and scarcity and inertia were actually net positive variables in the human lived experience.
This is not a blog post about “burnout,” because “burnout” is a symptom of hammering away at a single thing for an uncommon length of time. This is a blog post about the opposite: the AI-enabled compulsion to simultaneously hammer away at a million things for an uncommonly short time.
There’s a study in Barry Schwartz’s The Paradox of Choice that showed when a supermarket displayed 24 varieties of gourmet jam, 60% of shoppers stopped, but only 3% bought any. When they displayed only 6 varieties, fewer people stopped (40%), but 30% made a purchase.
What’s the lesson here? We suck at making decisions, particularly when there are several “good” choices.
Take dating apps as an example. Throw the entire meat market of eligible singles on the other side of a screen and no shit fertility rates are going sub-zero: as your number of perceived options approaches infinity, the “opportunity cost” of any particular option feels uncharacteristically high, which lowers the likelihood of committing to any one option, leading to a perpetual loop of infatuations that subsist until they are replaced by yet another infatuation, typically around the time that any real “friction” develops with your current infatuation.
60% of shoppers stopped (which, in our dating app instance, could mean anything from grabbed drinks to made out to had sex) but only 3% bought any (assumed any risk or vulnerability).
But this isn’t a blog post about sex, love, and situationships. This is a blog post about artificial intelligence, coding assistants, Claude, Grok Bot, Codex, tokens, harnesses, and “giving root access” to all of the above. Which is far more tantalizing than first dates, martinis, and the prospect of sex, right?
A Codex subscription is just Raya for folks who know what “reinforcement learning is” in that, instead of having your retinas bombarded with cleavage and witty prompts, your brain turns to the unlimited possibilities of what you “could” build.
“Programming” used to suck, because so much of “programming” was simply refactoring, debugging, and, really, just trying to understand what the hell you were reading in a particular Github repo. You couldn’t even “code” at the speed of typing (which, really, isn’t that fast), you could only “code” at the speed with which you could “comprehend” that which you were typing. Comprehension requires foundational knowledge, intuition refined through experience, and focus. Only so much (quality!) code could be written in so much time before you would have to take a break and resume your pursuit the following day.
This also meant that, obviously, there was a certain amount of “boot up” time required to context switch into a particular task. I have felt this with writing. When I was working on my book (if you’ve made it this far, buy your copy here), something I quickly became aware of was that I probably needed, at a minimum, ~3 hours of uninterrupted time to write or it wasn’t worth getting started. It would take me ~30 minutes to fully context switch and get back in the flow of where I left off, and maybe another ~30 minutes to “wind down” whatever section I was working through before I could actually step away.
I can’t tell you how often I labored over a particular word, or sentence, or figure of speech used within a monologue, just to delete, then rewrite, the whole sentence or paragraph or page or chapter before the puzzle finally fit together. You need big chunks of time where you can’t be bothered by calls or texts or emails to work through that puzzle.
For me, this meant that 95% of my writing happened between the time I got off work on Friday and went to bed on Sunday. I had more cognitive energy and free time on the weekends than I did during the week.
Requiring large, uninterrupted blocks on my calendar, while annoying in real-time (trust me, I would have much preferred to be able to write for 20 minutes here, 10 minutes there, over the course of the week without having to sacrifice the majority of my weekends in 2025) was actually, in hindsight, a good thing.
I knew that I had to write a book. It was an important personal goal of mine, and I had accepted a large advance that I was contractually obligated to deliver on. I also realized that the only way I was going to write this book was by aggressively protecting my calendar when I had those large blocks, only saying “yes” to important social events and otherwise ducking the fun-but-frivolous occurrences that would have been enjoyable in the moment but ultimately detracted from me accomplishing my goal.
The friction involved in writing was a necessary ingredient for me to construct my life in such a way to get the writing done. “The obstacle is the way,” as Ryan Holiday said.
There was no debating with myself, “Do I work on this book now?” When I had 20 minutes of free time on a Tuesday. “No, of course not. You won’t get anything done now. Save it for your actual ‘writing’ windows.” And then on the weekends, I would hold up the other side of the bargain: you’re not drinking this Saturday: you’re going to write 10,000 words. And so I did.
Good things take time, and when you recognize the time that they’ll take, you construct your life (and calendar) accordingly to position yourself to pursue those things. There are opportunity costs associated with anything that takes time, because, obviously, that time could always be spent doing other things, but once you internalize that long blocks of uninterrupted time are “necessary but not sufficient” for accomplishing anything impressive, it’s easier to make that trade.
If you want to write a book, you have to carve out the time to write a book. If you want to build software, you have to carve out the time to build software. If you want to pursue a relationship, you have to carve out the time to pursue a relationship. Any output worth having has historically required real investment as an input.
AI is changing that by bringing the “paradox of choice” to all of our white collar, digital work outputs. And I’m not so sure that this reduction in friction is a good thing.
My problem with our ever-improving AI tools is that they have decoupled the need for “focus” and “uninterrupted time” from the outputs previously associated with “focus” and “uninterrupted time.”
To be blunt, instead of having to sit down and “do” a thing, you can, while hungover in sweats on your couch, grab your laptop, fire up Claude Code, Codex, GrokBot, whatever, and just tell it to start building whatever epiphany is top of mind at the moment. Oh, and you don’t even have to type it anymore. Hands too sore? You can double tap the function key and your space bar, have Wispr Flow start transcribing, and say, “I want you to build a tool that each day goes through all newsletters I’m subscribed to, find the key highlights from each that I should care about, and send me a summary. Make this a daily routine, thanks.” Then go back to vegging out on your couch while your little token muncher cranks away.
Knowing programming languages is no longer a prerequisite for getting your computer to “do things” because AI can code better than any programmer who’s ever lived, and it happens to communicate with you and me quite well in English.
Do you realize how insane it is that we live in a world today where the speed between actively kicking off another project is effectively just opening your terminal, not even terminal, opening a desktop app, and just saying, “Hey Claude, you’re connected to all these things. Start building”?
At first glance, this is great! Software engineering, or, at least, “the ability to get a computer to do anything you want it to do,” has been fully democratized. That’s a wonderful thing. The issue, however, is that friction is a necessity for prioritization. When starting, and doing, a particular thing takes a certain amount of time, that friction forces prioritization. You can’t “do it all” therefore you have to consciously choose one pursuit over others. Choosing to spend 20 hours on a weekend working on a book meant not choosing 20 hours working on any other thing. There was a tradeoff, I understood the tradeoff, I chose accordingly.
Until approximately one year ago, starting “any” project involved some level of friction, but now you can spec out 10 different projects simultaneously by talking to your computer, and it certainly “feels” productive as all of your projects start churning away. But it’s not productive. You’re “doing work,” yes, but “work” has now been decoupled from focus.
You can still only turn your attention to a single thing at a single time. There has been study after study after study showing that we are awful multitaskers, while we at the same time have inflated beliefs in our ability to multitask. You think you can focus on multiple things at one time, but in reality you’re constantly just switching your focus back and forth to the detriment of both tasks, when you would have been better off just doubling down on a particular thing until you were done.
The friction of starting a particular task used to be enough of a hurdle to prevent us from getting in over our heads on doing too many things at once, but now we can “do” dozens of things at the same time and “feel” productive. But it’s like having dozens of browser tabs open at once on your computer: your own internal processing time begins to slow.
Oh, but you “can” do another thing. And another. In fact! Any time you have an idea that you could pursue, with the help of your omnipotent digital assistants, it would be an entire waste of your agency to not pursue it in your never-ending quest for more efficiency, or creating shareholder value, or whatever church you subconsciously subscribe to in our post-religious state.
“I need to automate my emails.”
“I need stats on my social media posts.”
“Have I used AI to aggregate every possible detail of this investment opportunity from every last corner of the web to compile into this memo?”
“What’s the best Greek spot in lower Manhattan for dinner? Which bar should I hit after? Can I automate the research for this?”
See, the biggest problem with capabilities only bound by computation and imagination is that, for the first time ever, we can pursue the marginal gain of anything and everything. And when there’s no limit on how many things you can hypothetically do (because the work can happen while you’re asleep, and there’s zero time needed to go from epiphany to action), you pursue each hypothetical immediately. Are there opportunity costs? Maybe, but they aren’t readily apparent.
Long onto Hinge. Swipe for 30 minutes. Check the app that night. 8 matches? You better go out with all of them, because why not? Don’t want to waste an opportunity.
Open your terminal. Fire up your bot. Redo your personal website. Have all of your newsletters summarized in one neat bulletin straight to your inbox. Who gives a damn about the source material, just give me the Spark Notes. Wire up an MCP server to my Spotify so GrokBot can change my playlist. Why not? Is my rolodex efficient enough? Can Claude update my Google Maps favorites? I’d love to sync it with Beli.
When you have an agent, everything looks like a token.
In Thinking Fast and Slow, Daniel Kahneman defines two types of thinking: system 1 thinking and system 2 thinking. System 1 thinking is unconscious, instantaneous, automatic, thinking. Recognizing a friend’s face. Reading a billboard. Reacting to a loud noise. System 2 is conscious, analytical, energy-intensive. It requires intentional focus. Filling out your tax forms (pre-AI!), analyzing a consequential legal contract, understanding the business model of a new company.
System 1 thinking is valuable because life would otherwise be unlivable without it. We need to pattern match and leverage heuristics to survive in the world. Imagine if you had to run an internal Monte Carlo every time you decided which sweatshirt to throw on or what to order at Chipotle. Just get the bowl with extra steak like you do every Thursday, it will be fine.
System 1 thinking is useful for helping us “do” most things. But System 2 is quite important for figuring out “what’s worth doing” in the first place.
The problem with our AI-driven world is that now, because the friction to taking action on, really, anything, is basically nonexistent, your first impulse is to simply begin doing the thing without first asking if you “should” be doing this thing in the first place.
“Busyness” begins to look quite a lot like “progress.”
Of course, there are ripple effects of this “do anything at first inkling of an idea to do thing” mindset: anything that does require System 2 thinking can feel like a “waste of time” because it prevents you from “starting” more, other, things.
Why read a full book when you can ask AI for the highlights? It’s the same thing, right? Do you really need to learn to program, or how code even “works” when you can just prompt your way to the right answers in 2026? Just get the answer. Do the work. Faster. Faster. FASTER. Anything that slows progress is to your detriment. You have the capability to do it all, don’t waste your time on any one thing.
Nasim Taleb, author of The Black Swan and Fooled by Randomness (among other books) wrote a fantastic blog recently called “The Book Is a Book, Not Disembodied Information.”
Basically, Taleb spent an essay discussing why Kindles suck, and physical books are the superior medium for actually consuming the contents of said book. To quote Taleb:
Believing that the book is just abstract information that can be merely held on a stick is an error we will call dimension overtruncation.
Definition: Dimension overtruncation is a Procrustean-bed in which the simplification of an object, phenomenon, or process proceeds beyond the level that can be justified without material loss. It applies to situations when the focus is confined to easily observable dimensions while less visible ones –while central –are ignored. Thus, an item may be assumed to be used for a stated purpose while its survival and persistence are attributable to other, unrecognized dimensions…
Things got worse when I met Daniel Kahneman on the following day and he was singing the praises of the Kindle. I tried to explain to him that reading virtual formats miss a fundamental dimension. I had just read the story by the Russian scientist A.R. Luria about some kind of memory genius whose ploy consisted in associating phrases with physical objects, applying the mnemonic trick ancient Greeks call the method of the loci. Each locus has to be spacially separated. Furthermore I had discovered quite early on that things you read online are hard to remember. Kahneman wavered between the excitement of neomania and the convenience of academic research on one hand and, on the other, a compelling psychological argument coming from his discipline. He agreed half heartedly. You could carry an entire library on that device and pick what you want on your next trip to Chicago, that type of thing, was his argument.
This is the classical Procrustean bed type of reduction in dimensionality. You think you are reading abstract content but we are not abstract beings and the physical book recruits more channels than you can intellectually perceive.
(Ignore the irony of Taleb calling out Kahneman, whose System 1 and System 2 thinking we’ve been discussing). Taleb’s point is that when you reduce a book to a sub-book form, it ceases to be that “book,” and it ceases to give you that which you should have received from that book.
Why do books matter?
Well, yes, you can find any singular answer, or series of answers, to any question(s) you like instantaneously with “AI,” but your learnings will be shallow. An author of a book has painstakingly presented the information in such a way that the readers can, through the course of the full book, gain either a foundational or deep understanding of the topic at hand (if the book is educational), or find themselves thoroughly entertained by the prose, assuming, of course, that the reader sticks along to let the passages develop.
Yes, of course, most books suck! And they shouldn’t be read! But that doesn’t mean we should try to expedite the learning of books worth reading. Summarizing a thing doesn’t give you the answers faster. It obfuscates the value of the thing itself, providing a perverted rendition that mimics the “thing” without being the “thing.”
Of course, we summarize in our bid to “save time,” because, again, when the speed of work is now instantaneous, anything requiring more than 30 minutes of your time is a waste of time. We have tokens to churn.
But time, focus, and concentration are prerequisites for discernment. How do you develop intuition? It’s not by hitting, cmd+c, cmd+v, and “enter” over and over again into a chat bot. It’s from doing the work, sitting with the work, and doing the work some more. The former just makes you a meat proxy spinning the slot machine waiting to see what token prediction Claude spits out on a Tuesday. You have to leverage system 2 thinking if you actually want to develop the capability for “critical thinking” yourself. That’s just how the human brain works (unless, of course, you prefer to be a meat proxy).
But because AI can give you an intelligent output more quickly than you can “do the work” to understand the output, you feel like you’re wasting time trying to understand any output. It’s on to the next prompt. But is efficient use of time actually the best use of time? I don’t think so. But we don’t give ourselves enough time to even contemplate the use of our time.
There’s this thought experiment that’s been echoed across the internet so many times over the last 20 years that it’s basically a meme at this point, but I’m going to retell it again here.
A professor goes to the front of his classroom and he has you pour a bunch of sand in a cup, and then you put some pebbles in the cup, and then you stack rocks in the cup. Oh no, the rocks don’t fit. The sand represents the “little things” in your life. The pebbles represent your “medium” responsibilities, and the rocks represent your “priorities.” If you put the little things first (doomscrolling on social media, wasting time on unimportant tasks at work, gossiping, etc.), you don’t have time for the more important things, and your life feels cluttered.
Then the professor grabs another cup and pours in first the larger rocks, then the pebbles, then the sand, and, “Wow!” everything fits. And he goes on to explain that you should first carve out time for what “really” matters like your health, family, friends, and career. Then the pebbles like your chores, cleaning your apartment, other extracurricular activities, etc. Then the nonessentials (sand) will always fill in the gaps around what’s left.
Of course, “token-maxxing” feels like a rock because it’s “work,” even though, if we’re being honest, most “AI usage” is closer to doomscrolling TikTok (when it comes to actual cognitive activity) than it is to “productive work.” And next thing you know, you have a cup full of sand that you could have sworn was rocks.
Social media isn’t the problem. Instant gratification is the problem. We just never had an instant gratification loop that fit so nicely into a “work” environment until now. Claude Code is just TikTok for folks who don’t think “sandboxing” is something you do on a playground.
The problem with “instant gratification” spreading its tentacles across all areas of our lives is that instant gratification leaves no room for compounding. We just keep starting thing, after thing, after thing, but none of those “things” stack upon each other. Motion without progress. Real knowledge, progress, discernment, value is downstream of committing real time and effort toward a particular endeavor. When you decouple output from effort, the actor is the one who suffers.
An example: it’s ironic that we have this “instant answer” machine and yet I struggle to remember a single thing I’ve learned from asking AI a question. To Taleb’s point, my recall of screen-related content is much worse than what I’ve read in physical books.
A couple of months ago I read 1873: The Rothschilds, the First Great Depression, and the Making of the Modern World by Liaquat Ahamed, and it was just stellar. Ahamed masterfully weaved a compelling narrative that covered the rise of the Rothschilds, the collapse of the Silver Standard (and its impact on Europe), the boom-bust-nature of financiers like Jay Cooke, and tales of stock market (and startup!) excess in Germany and Austria in the second half of the 19th century. But it’s not any particular “fact” from history that stood out as much as just developing a much stronger understanding of what this period of time was like for a particular group of people from a particular part of the global economy.
The beautiful thing about a good book is that you learn things that you never would have thought to look up yourself, because the author did the work of finding the “right” stories for you. They weaved together what may have otherwise been disconnected points to make them digestible for the reader.
You don’t get a foundational understanding of anything when you speedrun your search for hyper-specific answers. But reading the book is “a waste of time” when it takes away from everything else you “could be” doing, right? When do we ever have time to read when there are unlimited things that we could be “working on” 24/7?
The issue of course, is that everyone feels like everyone else is using this God-tier technology 24-7, imploring you to do so as well just to keep up.
If sales reps you’re competing against are using AI to find prospects, you should be doing the same, and doing it 24/7. If you’re a hedge fund analyst not leveraging 20 AI bots to find every last granular datapoint to inform how your stocks “could” perform and someone else is, you’re going to get smoked. You could apply the same logic to any domain: any time not spent leveraging commoditized, super-human “intelligence” to process as much data as humanly possible is time wasted. Right?
We’ve created a system where everyone feels like they have to be doing anything and everything just to keep up in an increasingly fast-moving market environment, and “keeping up with the current thing” is crowding out anything and everything else that might not have an immediate feedback loop.
This is just an accelerant of an existing work trend.
Before the pandemic, work typically stayed at the office. Investment bankers may have gotten their asses kicked from 9:30 AM to 1:00 AM, but when they went home, they were home. Then work-from-home started, and, “Oh! Now we can ping you 24/7 at home as well.” And now that work-from-home has been normalized, everyone is expected to be “on” at all times. There is no “logging off” of a demanding job. AI is a step-function accelerant in the “productivity-maxxing” lifestyle. Not only are you always reachable, you can now always be working, even when you, yourself, aren’t the one “doing the work.” But even if your bots are “doing the work” for you, you’re still thinking about that work, and that cognitive load has a cost.
What’s the point of this rant?
I mentioned earlier that this isn’t about “burnout,” it’s about the opposite: everyone is trying to do everything all the time, because, to an extent, we can pursue everything, all the time. But is it worth it? No one seems to be stopping to ask. It’s ironic that “AI” was supposed to free us of the shackles of “the grind,” automating the tedious tasks to free us up to pursue both higher-leverage and more satisfying pursuits. And yet, everyone is busier than ever.
There’s an unlimited amount of possible “work,” and the completion of work just begets and accelerates more “work.” By removing the friction to “work,” we didn’t absolve ourselves of work. We all became Sisyphus. Happy prompting.
- Jack
My first book, Young Money: A Field Guide to Wealth and Purpose in Your Twenties, is available everywhere you buy books. I wrote 270 pages on status games, opportunity costs, and the stage-specificity of life; topics that I think are particularly timely today. If you like this newsletter, you should get your copy below:
Find this article online or have the email forwarded to you? Receive my next piece in your inbox by subscribing here:
