AI: Age (of the) Individual
Use AI or get left behind
In today’s Diary:
Ask the lady on your phone
Use AI or get left behind
“Accenture” in a box
What’s the best AI today?
JUST DO IT
Dear Rational Optimist,
AI sent me into an existential crisis. When I realized I was talking to ChatGPT more than to my wife on weekdays, I had to ask: Was I outsourcing too much of my thinking? Or mastering the most powerful productivity tool ever?
This Diary is my follow-up to the Rational Optimist guide to AI, which generated over 500 email responses. Today in part 2, I’m writing with kids and young adults in mind, because AI will make or break their careers.
But first let’s talk about how AI impacts you.
It boils down to one word:
Agency.
In a world where most people are waiting for permission…
AI hands you the power to just do things.
What is agency? You know it when you see it.
Blake Scholl has agency. He was designing internet coupons for Groupon before he founded Boom to bring supersonic travel back to America.
Palmer Luckey has agency. He was creating virtual reality headsets before he founded Anduril, which is disrupting the military industrial complex.
Neither asked for permission. They just did it.
In a world where AI can produce PhD-level work, fancy schools and resumes no longer cut it. What matters most is the raw determination to make things happen.
Adults are used to thinking with limits:
“I don’t know how to code.”
“I’m not a designer.”
“I don’t have time to write a book.”
These are no longer valid excuses.
For twenty bucks a month, you get an army of tireless, smart, multilingual, design-savvy, code-capable interns. You can ask a question, generate research or teach your kid quantum mechanics, all in real time.
That’s why AI is underhyped. Anyone who thinks otherwise probably hasn’t used it or never got past the first version of ChatGPT.
“How are you doing all this so fast?”
I’ve been experimenting with AI like a madman. It’s supercharged my entire workflow. AI is now my writing buddy, research assistant and editor rolled into one. The Rational Optimist Society couldn’t exist without it.
Primarily, I use AI for learning and automating:
Learning. When you find yourself wondering something, whip out ChatGPT and ask. If any human has figured out the answer to your question, ChatGPT will know. This doesn’t mean wonder is dead—it’s just reserved for big philosophical questions.
My kids have learned this. If I hesitate for even for a second to answer a question, they say, “Ask the lady on your phone.” Now we have quick explanations for the orbital mechanics behind the Apollo missions or why octopi have nine brains. All without Googling and sifting through extraneous information.
You can use AI to help you learn anything. Take coding for example. ChatGPT is already a world class programmer. As of December, OpenAI o3 ranked as the 175th best programmer in the world. At this rate, it will soon be #1.
AI can teach you how to code or just do it for you.
“Claude, generate 10 ideas for an app aimed at finance professionals, pick the best one, build a playable prototype of that.”
I typed that in and less than five minutes later Claude spat out a prototype of a visualization tool showing the correlations between asset classes. Having never written a line of code before, I felt like a wizard.
You can also train AI to be an expert tutor for your kids. I had ChatGPT create an interactive game showing my six-year-old how the internet worked. That’s just one example—beats a dull textbook!
Automating. I interview a lot of people for the Rational Optimist podcast. The most time-consuming part of interview prep is watching “tape” on your guests. Now you can feed AI transcripts and ask it to pull out key points and quotes. Then it comes up with questions for you! I’ve lost count of how many hours this has saved me.
The more grunt work I can automate with AI, the more my brain is freed up for higher order thinking. It also allows me to do things I simply couldn’t before.
For years I’d toyed with creating an “Energy Abundance Index.” But I’d shelved the idea because stitching the data together looked like weeks of spreadsheet slog. One ChatGPT prompt changed that. In minutes AI vacuumed up data on fuel reserves, local output, imports and energy used per person—and mixed it all into a single score.
Now we have an ROS Energy Abundance Index. Yes, ChatGPT made the chart too.
What shelved ideas (or businesses) could you revive with a few clicks?
We’ve entered the age of the individual.
The dawn of the one man, $1 billion business.
There’s a new breed of startups where a handful of people leverage AI to do the work of hundreds:
AI image generator Midjourney has 40 employees and generates $500 million in revenue.
AI code editor Cursor rakes in $100 million in sales with a 20-person staff.
Blackwater founder Erik Prince told me drones make black swan events much more likely in war, as one guy can blow up a $10 million tank with a $500 flying IED.
AI does the same for business. A few high-agency individuals using AI can now compete against giant corporations. Suddenly you’re not one person anymore, limited by your personal experience and narrow expertise. You can be a researcher, writer, designer, coder and analyst, all at once.
How good is today’s best AI?
AI capability is doubling every 6–9 months. It’s not just getting smarter, it’s learning to reason, to see, to remember.
Google’s DeepMind models are now competitive at the International Mathematical Olympiad level.
There are likely fewer than 200 people in the world who could beat OpenAI’s o3 in a competitive programming matchup.
The Graduate-Level Google-Proof Q&A (GPQA) is a benchmark of multiple-choice questions designed to test reasoning in advanced academic topics. Human PhDs, even with internet access, score around 81% on questions within their specialty.
Today’s AI wipes the floor with them.
Remember, this is the worst AI will ever be.
New AI models are constantly leapfrogging each other.
So, I hesitate to recommend one over another. But you can’t go wrong starting with the new ChatGPT-4o. It’s not the best researcher or coder, but it’s your reliable all-arounder.
While you’re at it, try ChatGPT’s new Deep Research feature. Think of it as a consultant in a box. You type in a question like “What’s the latest in battery tech?” and hit the Deep Research button. Then the tool scans articles, PDFs, websites and any files you upload. A few minutes later, you’ve got a crisp, well-organized research report.
It’s impressive. It feels like you’ve sent a research assistant off for two weeks and gotten back their best work.
AI won’t steal all the jobs. But work will evolve.
My neighbor works for Accenture, the $175 billion consulting giant. His job goes like this: Some organization calls up Accenture with a question. He does a deep dive into the problem, reading a bunch of stuff and talking to people. Then he hands a research report to the client six weeks later.
Consulting is basically college assignments for grown-ups. This type of work is at extremely high risk of imminent disruption.
So, how do we compete with AI? We don’t. Instead, we must…
Be more human.
AI will lead to unprecedented prosperity. But like every technological revolution, there will be winners and losers. Where you wash out is up to you. See: Agency.
Venture capitalist Josh Wolfe says, “What’s scarce is valuable.” What’s scarce in the age of AI abundance? Your authentic humanness.
The paradox of AI is the better it gets, the more valuable our stories, quirks, tastes and weirdness become.
If you compete against a computer, you will likely lose. But if you partner with a computer, you can win big. Your job is to figure out where the machines excel. Then let the computer do its thing while you do yours.
Think of it as a barbell strategy. On one end, you must be a hyper user of AI. Use the machine for research, data crunching and coding. Load the other end of the barbell with YOU. Your unique perspective and worldview. Thanks for the visual, ChatGPT.
Be more human; that’s one job AI won’t replace. Those quirky, messy, soulful traits that machines can’t replicate are your new superpower. In practice this means:
Relationships run the world. AI can’t share a genuine laugh over coffee, intuit what someone needs before they ask or build deep trust through shared struggle. The ability to connect with other humans is rocket fuel in the AI age.
“Networking” goes from a useless LinkedIn game to the #1 thing you need to succeed. I use AI to get things done so I have more time to meet people face to face. Yes, I used AI to do a first edit of this essay so I had more time to prepare for my talk with Matt Ridley.
As an aside, I’m bearish on AI doing original writing. AI has no soul, and neither does the writing it produces. AI will never replace good writers – it will enhance them!
Use your imagination. ChatGPT doesn’t have urges, dreams, or spontaneous creativity. It doesn’t wake up at 3am with a breakthrough idea. The future belongs to those who can imagine what doesn’t yet exist and use AI to bring it to life.
Be illegible. AI thrives on clear rules and measurable outcomes. That’s why it's blowing past humans in math and coding – where right and wrong outcomes are clear and you can train a model until it masters the game.
If your work can be clearly graded, AI will eventually do it better, faster, and cheaper. But AI cannot “understand” how a jazz musician chooses the perfect note. Or how an investor knows exactly when to break his own rules. The magic happens in the weird connections only your brain makes.
Your weirdness and unusual mix of knowledge is now your moat. “Illegibility” is what makes you valuable in the AI era. Time to double down on being unmistakably you.
AI is your big opportunity.
A few thoughts for two important groups:
Entrepreneurs. The Age of the Individual brings incredible new opportunities. YouTube creator MrBeast, with a core team you could fit inside a school bus, made a sports video with more views than the Super Bowl.
The internet birthed dorm room billionaires. Now AI is further amplifying individual human potential. We’re entering an era were getting rich from your passion isn’t just possible, it’s probable for those with agency.
I guarantee many people will build life-changing wealth by leveraging these tools. Go be one of them.
Parents (and grandparents). If you have kids or grandkids, you have a duty to prepare them for what’s coming. Don’t wait for your school to “adopt” AI. Your kids already have access to a world-class, infinitely patient tutor for every subject. Get ahead of the trend by carving out 30 minutes a week for AI practice.
Smartphones and social media are mind viruses for kids. But AI is the polar opposite of TikTok. It’s not mindless dopamine scrolling. It demands your input and curiosity – a huge shift in the right direction.
“AI doesn’t apply to me.”
I promise you, it does.
AI is already transforming law, medicine, education, investing and writing. Future winners won't necessarily be those with the fanciest degrees or most experience. They’ll be the high-agency individuals.
To quote ROS honoree Peter Thiel:
At the extreme, optimism and pessimism are the same thing. If you’re extremely pessimistic, there’s nothing you can do. If you’re extremely optimistic, there’s nothing you need to do. Both extreme optimism and extreme pessimism converge on laziness. I believe in human agency and that it's up to us.
Don’t wait until you understand AI to try it. Simply fire up ChatGPT and ask it to help you with something you already do. You have been given a new superpower. Use AI or get left behind.
And please, don’t ask for permission. Or as they say at Nike, JUST DO IT.
Want to help?
Forward this email to all the young people in your life.
Then, hit “Like” and “Restack” below to spread the word.
Future’s bright when you have high agency!
Stephen McBride is a co-founder of the Rational Optimist Society.
Hi Stephen,
I have to disagree with you that AI is underhyped for the same reason Alphabet executives accepted resignations of AI researchers rather than release a product they felt wasn't ready. LLMs still average a 15% error rate. Use of such faulty products in commercial settings have already resulted in money losses and law suits.
Your recommended personal uses still need to come with cautions, which I find lacking in your articles, the muted comments at the end notwithstanding. I ran into an embarrassing use case last week that “trust, but verify” also applies to learning.
Absolutely, LLMs are a great learning tool, but you need to also tell people that they need to check the answer against the listed or other sources before using them in work or homework. LLMs don't understand English. In my tests, the error rate was as high as 100%, If you grade like an English teacher who doesn't allow even a single mistake (Need I mention that my mother taught high school English?).
Every national news outlet I watched mangled the name of the German chancellor. I'm sure they asked chatbots in English, because when I did so, they all made the same mistake. Why? Because they reported the most frequent English answer, which is how people who anglicized the name pronounced it. As is common in French or English, they softened the Z to an S, and some also soften the E. You can't do that in German, which has strict pronunciation rules. To find the right answer, though, I had to ask in German, which produced a direct hit of a German speaker pronouncing “Friedrich Merz.” He was probably equally frustrated about the answers from native English speaker, but he didn't have the reach to win the AI probability race.
Bad enough that national news outlets embarrassed themselves. From an educator’s point of view, this is terrible, because they spread misinformation across the country that now needs to be rooted out, which is probably impossible. I've found it at best a 50-50 proposition to point out to people that corporate videos and Jensen Huang himself pronounce the name “Envidia.”
So, please, mention at least once in each article that readers should check answers against source material before relying on them for public pronunciations or anything more serious than that. Which, by the way, is agency.
And please stop relying on the one use case that proves the rule. I absolutely agree with you that chatbots write better code. There's a good reason all major software out there is so bad. Most humans can't think like a computer. The big exception is high-functioning autistics who think like Spock. For good reason neural dirvergents are over-represented in programming. In contrast, AI thinks like a computer because it is based on a computer. You can't extrapolate from that to cases involving the highly illogical English language. Ask Spock to answer “How do you feel?” and see what answer you get (See Star Trek IV The Voyage Home)
As an aside, please also stop doing in reverse what you chide others for who’ve only used the first version of ChatGPT. You obviously haven't used Google lately. I've stopped loading Gemini, because Google by now knows when AI is the better answer, which is almost always. Here’s the beginning of the answer to the query you mentioned in the article.
“AI Overview
The Apollo missions relied on a series of carefully planned orbital maneuvers, including transfers, rendezvous, and orbit insertion, to achieve their objectives of landing on the Moon and returning to Earth. The Hohmann transfer, a technique for changing orbits, was a key element in these maneuvers, as were retrograde and prograde burns to adjust the spacecraft's altitude.
Here's a more detailed look at the orbital mechanics:
1. Launch and Earth Orbit
…”
How "public" is the trail I leave with Chatgpt? Am I basically publishing my logic stream about my new business idea as I interact with AI about it? Does Chatgpt use what it learns from the questions I ask and just give it to others asking similar questions? It would seem that going forward claiming that something was actually my idea becomes fuzzy with this, and keeping any control over it for personal profit quite problematic.
Maybe a whole lot more sharing of ideas is a good thing. This looks like a lot more with no control over who you share yours with.