The trillion-dollar perception gap between Silicon Valley and normal people.

Maybe the real AI winners are the folks who just don't care?

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.

---

Brief note: I’m co-hosting a San Francisco Tech Breakfast Club next Tuesday from 9-11 AM PDT. If you’re around, come hang, caffeinate, and chow down with friends. Signup link here.

I spent most of 2025 living in San Francisco.

Before December 2024, I had never visited San Francisco. In fact, my only experiences on the west coast before December 2024 were a trip to San Diego with my grandparents for my grandpa’s board meeting when I was 12, a trip to Los Angeles for the Rose Bowl in 2017 with my then-college girlfriend, in which we slept in an RV camper in a tailgate parking lot with her family (they lived in LA, but went all-out for the Rose Bowl every year. This was when UGA pulled off a miracle over the Baker Mayfield-led Oklahoma. Sick game. I think that ex-girlfriend is engaged or married now. I still have my Hinge radius set on four miles of Union Square, Manhattan and filtering for 20-something chicks that are Christian and/or Catholic. Who’s the real winner?), and a wealth management conference called “Future Proof” where I saw Third Eye Blind play live in September 2024.

I had never visited San Francisco.

Then I went out there for a job interview. Self-driving had always “sounded cool” until I flew out there and tried my first Waymo. I’ve never done crack cocaine before, but I imagine the “wow” factor is similar. That thing blew my mind. Car driving itself. Insane. Suddenly, I could connect my Spotify to the surround sound in a fancy Jaguar and jam to whatever I wanted without some stranger with a 4.88 Uber rating who smelled like a mixture of cigarettes and stale pizza judging my music or eavesdropping on my phone calls. I think there are a list of “firsts” that are quite memorable for dudes born in the 90s.

• First touchdown (or goal, homerun, etc).

• First time bench pressing two plates.

• First time sipping alcohol.

• First time having sex.

• First time beating the last mission on Halo 3 on veteran difficulty.

• First time riding in a self-driving car.

On the second point, if you’ve never benched two plates, I think it’s hard to call yourself a dude. Especially now that everyone is ripping peptides. Get your weight up. I digress.

Anyway, the future is now in San Francisco. So I ended up moving out there for a year. Where I’m from, in Georgia, “self-driving” is some mythical future from the Jetsons. Except in Atlanta, where sometimes the Uber you order is a Waymo, and sometimes it’s some stranger who doesn’t match the Uber picture, and their best friend in the front seat with blunt wrappers on the floor, and you wonder “why did I pay $20 for this?" For some reason, Atlanta forces you to order Waymos through Uber rather than just using the Waymo app.

Anyway, self-driving is this insane, futuristic-but-not-realistic thing in most of America. In San Francisco, it is my default form of transportation.

Oh, have you seen the housing market going crazy in the Bay Area? It’s because OpenAI just raised like $120 billion at an $850 billion valuation in preparation of their $1 trillion IPO despite the fact that they’re about to incinerate another $100 billion+ on data centers. Does the business work? Who cares. As long as Masayoshi Son will buy $20 million+ of OpenAI stock from AI researchers (read: folks who figured out how to get “next token prediction” to convince you that, yes, your idea which isn’t great is, actually, quite genius), and Nvidia continues its generational run of printing infinite money by selling their former bitcoin mining rigs to companies who want to generate infinite tokens by forcing their developers to vibe-code 100,000 lines of code per day, it’s going to cost you $5 million for a 3 bed, 2 bath home in Pac Heights.

I set up an old Mac Mini a month ago.

Why? Well, when I got on Twitter in February, I saw that some Austrian / British dude named “Peter” built an AI assistant that you can tell, through Whatsapp, to order pizzas for you. If you’re not careful, someone will prompt inject your OpenClaw and steal your social security number and the nudes your girlfriend sent you, but hey at least it can monitor your email for you. OpenAI “acquired” Peter (but not OpenClaw, which is a nonprofit foundation, similar to how OpenAI was originally, before it became a trillion-dollar, for-profit company. Because who the hell wants to be a non-profit when Masayoshi Son will shove $100 billion down your throat and speedrun WeWork again?) for about as much money as they paid to acquire a niche, but popular, daily tech show: hundreds of millions of dollars.

There’s this meme of “escaping the permanent underclass.”

That meme basically means you have about 18 months left (probably less, now) until AI gets so good that money, and really your entire existence, is rendered totally irrelevant. Which is a hysterical conclusion to draw from a “next token prediction” machine, regardless of how good it is at predicting tokens, but it seems like everyone in San Francisco believes it. Either A) you better get hired by an AI lab, or B) you better raise $100m for your startup and get acquired by an AI lab, by 2027, or… adios! Off to the token mines for you. Go make some human training slop to create marginal improvements for Opus 5.69 or whatever.

Of course, maybe there is some truth to the “permanent underclass.” At least in the engineering world. See, LLMs can’t replace most jobs, in my opinion. Most jobs have some level of nuance or friction, and LLMs are, really, word-and-number prediction machines. They’re quite good at these predictions, but, again, they’re kind of limited to screens. But coding? Well, coding is just a bunch of words and numbers, and it’s easily verifiable. The code either works, or it doesn’t. Give an AI enough context and a big enough context window, and it can build basically “any code.” And San Francisco, which was originally a gold rush city, has made a lot of people who knew how to code rich over the last 30 years. New platform comes out (computers, the internet, mobile), and the folks who could “use code” to build on that platform (Facebook, Airbnb, Uber) made a shitton of money. But now, “code” is commoditized. The biggest value prop in “tech,” being the guy who can build the digital money printing machine, went from a tightly held secret to accessible to anyone with a keyboard and 5th grade reading comprehension.

No wonder everyone in the Bay is so anxious.

Also, shout out to Stack Overflow for selling for $1.8 billion just one year before ChatGPT rendered it redundant. Talk about securing the bag. That feels like a microcosm of what’s happening in tech in general: get your bag before you get rug-pulled. And the timeline for securing that bag keeps shortening.

Anthropic now has a “Mythos” model that found several security holes in the Linux, and helped patch a 27-year-old bug in OpenBSD, an operating system, two years older than me (launched in 1995), known for its security. This thing is so powerful that Anthropic wouldn’t even launch it. They just tweeted about it, and said that 12 companies have access to it, and it’s too dangerous to release publicly right now because everyone and their mom could use Mythos to hack everyone else and everyone else’s moms. That’s pretty good marketing, to be fair. “Hey, we build AI God. It’s super scary. We’re telling you about it, but you can’t use it. Sorry!”

With Mythos, you could, hypothetically, say, “Hey Claude, steal all of _____ internal secrets. LOL what a good prank.” and it could probably do that?

The marginal value of the marginal software engineer is quickly plummeting to zero, and a small group of people and companies with GOATed models are operating on God-mode right now. It’s like a zero-sum race to extract the last amount of money left while money is still a thing.

Anyway, I moved back to New York two months ago. A couple of my buddies asked me, “Have you used Perplexity?” I went on a date with a girl who “worked in consulting,” and you would be shocked, SHOCKED, to learn that not once did she ask me about what harness my OpenClaw runs on. I think most of my friends are using ChatGPT at this point (?), a few know what “Claude” is, no one has touched a terminal, and the word “bash” sounds a bit like what a Neanderthal does with its club. I’d like to see Opus-4.6 split the G, or rip a Citibike across the Williamsburg bridge. I went to a Broadway show a few weeks ago. No AGI in sight.

What’s the point of this ramble?

The disconnect between the epicenter of AI (San Francisco, and. more broadly, a certain subsection of Twitter) and the rest of the world is just wild. The former truly believes we’re on the verge of a world-changing paradigm shift. The latter just doesn’t care. Perception doesn’t just shape reality; it is reality. And, really, I think both can be true. In a world that lives and dies by the power of bits, well, yeah, an automatic bit generator certainly changes everything, doesn’t it?

But in a world that just wants its Uber to arrive on time, (or, even better, a Waymo with no mystery man behind the wheel), which is most of the “real world,” by the way, some super-powered coding assistant is just, like, irrelevant. And that’s what’s funny about this whole AI boom: most people don’t care. And, hysterically, they don’t have to care. I mean, like, yes, they could, hypothetically, get a lot of leverage from all of this stuff. But does it “really” matter? I’m honestly not sure.

Again, I subscribe to the belief that AI can, and will, automate all code. “Coding” just becomes another form of writing. Some people will be prolific, others will suck, but everyone “can” do it. Or soon will be able to. Guess what? The invention of writing was pretty cool. We’re all just used to it now.

I don’t think “AI” will impact most other jobs, occupations, and sectors, nearly as much as “code generation” because there are too many edge cases. So, I guess, I see the world going something like this:

• The speed in which better and better software is released and updated will be astounding.

• 99% of people who have never written a line of code will never write a line of code.

• That 99% of people will benefit, as consumers, from cheaper code (competition drives prices to zero when anyone can generate code for no cost!) by having access to an ever-growing selection of impressive software. They might notice, “Oh wow, my apps don’t glitch as much,” but that’s about it.

• That 99% of people will remain totally unaware of the existential-levels of anxiety happening in the Bay, and they’re probably better off because of it.

On this last point: I feel the angst. I have, somewhat, sipped the “AGI kool-aid.” I panic-forced myself to figure out how to use Claude Code to build different stuff because. I thought it was going to be table stakes to “survive” as technology continues to accelerate. I’ve used Claude and ChatGPT and Grok and Perplexity and Wispr and Parallel and Exa and Manus and… you get the idea. I have a personal AI assistant named “Clank,” that I can talk to through Telegram, who can order pizzas for me or figure out who, in my rolodex, I should grab beers with next week.

And yet, all of my friends who have no idea what the hell a “Command Line Interface” is are making fine money living in their Manhattan shoe boxes, just like me.

There’s a tale of two worlds happening right now: a small subset of people speedrunning “AGI,” and then everyone else, who couldn’t care less. The former might make a lot of money, at least in the short-run. But the latter is probably having more fun.

Ignorance is bliss. Enjoy the permanent underclass, it’s fun down here.

- 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!

---