A week or so ago I had the opportunity to speak to some high school kids as part of my friend Makiko’s program. I always love talking with them. They are super smart, don’t let things slide, and always push me.

Often when I am in these situations I go into a fugue state and whatever I have been thinking about for the last couple weeks surfaces. A severe case of recency bias.

The session seemed to resonate. Afterwards, I got curious about exactly what I had said. Weird, I know.

So, like any good AI-enabled human, I passed the transcript to ChatGPT 5.6 Pro and was like, “how did it go?”

The results were interesting. I liked what I said!

And, reading through them, I realized this is basically what has been bouncing around my brain lately. So I thought it was worth sharing with a broader audience.

This is the edited annotated response from ChatGPT :

ChatGPT is block quoted

The strongest gems

“The candidate is always greater than the technology.”

This is probably the cleanest Obama-era lesson in the whole conversation. Tech is a force multiplier, not the underlying force. You immediately generalize it too: great tools don’t rescue something people fundamentally don’t want.

I used to get this question constantly: “would the work y’all did for Obama work for candidate X,Y or Z?“

I don’t think it worked like that in 2012 (our tech would never have worked for Romney, and I suspect it doesn’t work like this in general) - you can’t make a bad candidate better with superior technology. Elections are won by the candidates, not the tech stack. I think

“The Narwhal misinformation loop is an incredible story.”

The press misunderstood what Narwhal did → the mythology became accepted as fact → Republicans and Clinton-world studied that mythology → they actually built the imaginary system.

Misinformation can become true when people act on it.

Much more interesting than the usual “fake news is bad” conversation.

Narwhal is such a funny thing. It was effectively a way for OFA in 2012 to have a shared login, identity and user profiles. It wasn’t really used for targeting at all. The resulting narrative about what Narwhal did and the impacts of Narwhal have all been overblown.

“Silicon Valley doesn’t actually want you to be an entrepreneur. It wants you to become a worker.”

Your argument: the ecosystem celebrates founders, but structurally benefits from thousands of ambitious people moving there, failing, and becoming employees of incumbent tech companies.

Not everyone can be an entrepreneur. It is a convenient pipeline to tech worker. High Agency! Self Starter! Had a startup before!

Always ask: “Who has power in this system?”

This appeared almost incidentally during the Silicon Valley discussion, but it’s one of the best general-purpose analytical frameworks in the conversation:

When someone tells you how a system works, ask who benefits from you believing that version.

I think one of the most important things a person can do is to interrogate who has power in a particular situation, system, etc.

AI changes the career ladder more than it eliminates the bottom rung.

Your contrarian argument was stronger than “AI takes junior jobs”:

AI may make a talented beginner dramatically more capable, which threatens expensive mid-career expertise—but then creates a new problem: how does the beginner become senior?

That career-ladder question is much more interesting than “will AI replace programmers?”

I really believe this. There will be a correction once the CFOs realize that a junior that costs $120k is equivalent to a mid stage employee who costs $350k. I worry a LOT about how a career happens in tech in about ten years.

Abundance vs. efficiency.

If AI turns $10M of productive capacity into $5M, the boring response is to spend $5M and fire half the people. The interesting response is to keep spending $10M and produce vastly more.

That is basically an entire 2389 thesis in miniature: AI should create abundance, not merely headcount reduction.

We will look back at this time period and laugh at all the companies who are using AI to make things MUCH MORE EFFICIENT. They will seem lazy and boring. The winners will be the companies that do so much more than they are currently doing. Why wouldn’t you?

“Authenticity is often linked to privilege.”

Excellent correction to the usual TED-talk advice to “just be yourself.” You explicitly point out that your ability to walk into investor rooms dressed however you want is partly enabled by how society reads you.

Authenticity is easiest when conformity is optional.

I think about this every time I look like an idiot and are in a “special” type of place. There is a special form of white privilege that allows people like me to look like idiots and fit into business contexts.

Diversity becomes more important when AI commoditizes execution.

If everybody can build roughly the same thing, differentiation increasingly comes from the context, culture, taste, experiences, and values people put into it.

In other words, when technical capability becomes cheap, human variance becomes valuable.

We see this every day at 2389. The more diverse team we work with, the better the output. Everyone talks about taste, etc - but nobody talks about how to stack the deck to enable taste. It is clear that having a diverse team is one of the ways to stack the deck.

“Fill your brain with ceramics.” 🏺

Perhaps the funniest formulation of a genuinely serious educational philosophy. If machines increasingly absorb procedural knowledge, don’t spend your entire education memorizing procedures machines are exceptionally good at. Develop taste, context, culture, judgment, physical skills, languages, music, art—and acquire technical fluency through praxis.

Learn the things that make you interesting; use the machine for the things that make you productive.

I recently learned the word praxis from my friend Dirk. It feels like there is a lot of opportunity to learn complex topics by working with AI to do work. Also, why learn to program computers (or another technical skill that AI can do really well) when you could learn to do cool art shit AND program.

Entrepreneurship as love rather than rational optimization.

Startups have terrible odds, yet probability alone doesn’t explain why people start them—just as failure statistics don’t explain love.

“Entrepreneurship is really about faith. It’s really about belief. Do you fall in love with this thing?”

“Think things logically, then make decisions emotionally.”

Taste isn’t logical. Romance is important. You MUST fall in love.

“If someone says you can’t do something, that’s information about them, not information about you.”

Very sticky. It transforms discouragement from a verdict into data about the speaker.

Everything is data. The data around you isn’t you. It is just data around you. Learn from it, but don’t let it define you. You can do anything.

The variance idea connects photography, engineering, and entrepreneurship.

“As you increase variance, you often increase the mean.” - Vaughn Tan and Mako Hill at Foo Camp

Take more photographs. Try more approaches. Start more things. Increase the number and variety of attempts and you increase your opportunity to encounter an exceptional result.

I am a horrible photographer, but I make up for it by taking a fuck ton of photos. Practice, praxis, high variance seems to work for me. I think this probably works for a lot of things.

Fun isn’t a reward for finishing life correctly.

Your rejection of “what did having fun contribute to your career?” was good precisely because you challenged the premise:

Why does fun need to justify itself economically?

And then you nevertheless showed that clubbing taught you networking, reading rooms, approaching strangers, social confidence, and connection. The instrumental benefit came because you weren’t optimizing for the instrumental benefit.

I spent a lot of time in clubs in my 20s and 30s. This was instrumental in defining my career outside of that space. I think we are all trying to work too much. Stop it. I regularly tell people around me that they shouldn’t work so hard when they are young. Being young in a city is made for fucking off. go fuck off. If you are not young, then fuck off anyway.

Vulnerability as infrastructure for trust.

You framed vulnerability less as emotional self-expression and more as a mechanism:

vulnerability → connection → trust → collaboration.

That’s a useful leadership formulation, especially alongside your “connecting people” superpower.

I told the class that I have a superpower for connecting people. I recently was part of a really interesting program (LIT) and it talked a lot about vulnerability in a very systematic way. I am obviously reflecting some of that training - but it fits and is important.

The larger worldview underneath all of it

What surprised me is how coherent the conversation is despite being extremely free-form. You keep returning to essentially five beliefs:

• Agency: just do the thing; permission is mostly imaginary.
• Iteration: more attempts, more variance, more chances to find something exceptional.
• Humanity: taste, fun, culture, vulnerability, relationships, and identity matter more as machines get better.
• Power: interrogate who benefits from systems, narratives, and supposedly inevitable outcomes.
• Abundance: technology should let humans do more interesting things, not merely allow corporations to employ fewer humans.

That feels like the actual payload of the whole session.

And maybe the most Harper line in 10,000 words is still:

“Put the things in your head that you value, the things that you love.”

Everything else—AI, programming, companies—is downstream of that.

Nailed it. Come hang out