The Future

Predicting the future is a fool’s errand. Almost everyone gets it wrong, especially the people most confident that they won’t.

I was listening to the How I Built This podcast with Reed Hastings this week. In 1997, when Netflix started, VHS tapes were still king. The only reason they used DVDs was practical, not visionary—they were lighter and cheaper to mail. That’s it. No grand theory about streaming dominance. Just postage. Netflix nearly died during the dot-com collapse in 2000, pivoted repeatedly, survived mostly by accident, avoided merging with Blockbuster because they were myopic, and eventually became a verb. Nobody in 1997 would have drawn a straight line from padded envelopes to global content dominance.

But that’s how the future actually works. Amazon. SpaceX. And now OpenAI and Anthropic. Not clean arcs—very messy iterations. When we started OneEnergy we knew one thing: don’t go to the Permian, it’s crowded…. New Mexico worked out pretty ok as it turned out.

Today, I look at AI through two lenses. One is as a law student, watching the entire profession slowly (too slowly) realize that a large portion of its high priced work can now be done faster, cheaper, and better by a machine. The other is as a father of two young men—21 and 19—who are about to compete for entry-level jobs against a robot that is, frankly, smarter than anyone we know.

We now carry more computing power in our pockets than existed on the spacecraft that allegedly—if you enjoy moon-landing conspiracy humor—landed on the moon in 1969. The fact that we “can’t” do that today says less about our ability to design and more about how radically our safety standards have changed. I have no doubt Elon will eventually put a small army of Optimus robots on the moon to build some forward city we’re told we could cohabitate in. I have no idea why anyone would want to live on the moon, but that’s beside the point. As a SpaceX shareholder, I just hope their stock goes up!

Which brings me to the present. My 21-year-old recently discovered Claude Code from Anthropic. Personally, after a close friend recommended I try it, I think Claude is meaningfully better right now than Grok, Gemini, or ChatGPT—but that’s a snapshot, not a prophecy. The larger point is that we’re watching trillion-dollar companies with effectively unlimited capital shove the frontier forward by brute force. That alone should concern anyone who thinks their existing moat is permanent. AI won’t take jobs, per se. But it will take jobs from those not using it. Because, what matters more is what happened this week. Ben learned to code.

Not “learned to code” like we used to mean—took classes, struggled through syntax, maybe built something six months later. He learned to code in about a day. Within six days, he had designed a game. He wrote a stock-screening algorithm that runs every morning. It ingests data, filters it, flags opportunities, and executes tasks without supervision. It’s… good. Uncomfortably good. I know because he calls me six times a day to tell me what he found.

Five years ago, the Biden administration gave this same advice to oil and gas workers: learn to code. The difference today is, with AI as a co-pilot, learning to code is no longer a career path—it’s a weekend decision. When someone can go from zero experience to writing functional applications in under a week, that’s not a productivity gain. That’s an economic regime change.

So what does that mean for the future?

Maybe entry-level workers will no longer be “entry-level” in the traditional sense. They’ll arrive with real systems already built, ideas already tested, tools already deployed. Maybe they’ll surpass their bosses knowledge in weeks. Or maybe they won’t seek jobs from institutions at all. Maybe they’ll trade the stock market themselves—Wall Street professionals—moving their own money or their grandparents money with the great wealth transfer happening and outperform “a career.”

As I’ve mentioned, the boys run their “hedge fund.” So far, they’ve massively outperformed the market. I don’t mean a lucky trade or two—I mean sustained performance over nearly a year. They’re finding companies I’ve never heard of, pulling data I wouldn’t think to pull, and identifying trends early. Some of those positions are up 50%, 60%, 70%. That’s not normal. And it’s not magic. It’s leverage—human judgment amplified by machines.

This is a brave new world.

If you haven’t tried Claude Code, it’s kind of unbelievable. And if it makes you uncomfortable that my AI skills have already been surpassed by my son—six days in—good. That discomfort is the signal. It’s the same signal Blockbuster missed. The same signal newspapers ignored. The same signal every disrupted industry shrugs off right before it’s too late.

The future doesn’t arrive fully formed. It shows up as a toy. Then a tool. Then a threat. Then a fact. We’re somewhere between tool and threat.

And the kids already know how to use it.