What I love about AI

ai

Let me tell you about Sergio, and why I’m incredibly optimistic about AI. But first, one of the things I really hate: the homogenization of everything. It doesn’t matter where you go — every strip mall is the same five tenants, every main street the same chains. Once in a while I find a gem — a typewriter store, a tiny bookstore with one specific aesthetic, a shop that exists because one person with taste decided it should. And every time, I think: why don’t we have more of these?

Sergio’s school

Sergio is not a curriculum architect, a state-compliance lawyer, a marketing-and-admissions person, a financial-aid administrator, or a development director. He’s a guy who knows the kids in his neighborhood and has a clear vision of how they should be taught — project-based, his own pedagogy, his own curriculum. Until very recently, opening a small school meant either paying five professionals their hourly rate or knowing five professionals who’d do you a favor. If you had neither the money nor the network, you didn’t open the school. You worked at someone else’s.

Now Sergio sits down at a laptop and gets to work.

He drafts curriculum units himself with an AI walking him through state standards, flagging where his project-based approach hits required outcomes and where it needs a tweak to comply. He generates a parent-facing prospectus in his actual voice — not the generic one a marketing firm would have written — and iterates it until it sounds like him. He pulls tuition comps for the region and gets a one-page operating cost model with realistic teacher salaries, lease ranges, and a break-even count. He drafts the 501(c)(3) paperwork, builds a teacher hiring rubric tuned for the kind of educators his pedagogy actually needs, and gets a checklist of permits with the city’s specific forms pre-filled.

None of this replaces the compliance lawyer, the curriculum consultant, the development director, or the admissions person. He still hires them where it matters. But he walks into each conversation 80% of the way there. He knows what the standards alignment looks like. He knows the tuition is in range. He knows the prospectus says what he means. The professionals get to do the genuinely hard parts and Sergio doesn’t get fleeced on the easy ones.

That’s one school. Sergio’s school.

Sergio gets his school. So does the next person.

The doomer narrative is loud. The optimism case is louder.

If you read the news, AI is going to take your job, flatten culture, make everyone dumber, and replace your friends with sycophantic chatbots. I’m not saying none of those concerns are real. I’m saying they’re getting the disproportionate share of the oxygen, and the actual things AI is doing for actual people — right now, in the worlds I touch — are extraordinary and underreported.

Three threads I keep pulling on: AI helps you think, AI lets you create, and AI clears the tax on starting things small. Each one is more important than it sounds.

AI helps you think and learn

The version of “AI for thinking” that gets you in trouble is AI for outsourcing thinking — let the model do the cognitive work, take the output, ship it. That’s not what I mean.

What I mean is the hyper-personal version. The same idea explained in three different ways until one of them lands for the specific person reading it. The math problem unpacked at exactly the level you’re stuck on. The framework re-cast through the metaphor that works for your brain, not a generic one. A tutor who knows the specific prior misconception that’s blocking you — not the misconception in the textbook, the one in your head — and aims directly at it.

I’ve watched this happen with my own kids. I’ve watched it happen with myself learning new code. I’ve watched it with people who’d given up on a topic because every introduction they ever read assumed background they didn’t have. AI doesn’t assume. AI asks. And then AI explains, and re-explains, and re-explains, and never gets bored.

The deeper version of this is the AI second brain. It’s not just that AI is patient and personalized — it’s that AI knows you. Your goals, your history, your patterns, the thing you said in a journal entry eight months ago that contradicts what you’re saying today. A coach who has read every page of your life and forgotten nothing. That’s the thing I’m chasing in my own setup, and the closer I get, the more I notice: when AI has rich context about you, everything it does becomes hyper-personalized by default. The advice. The reading list. The questions it asks back.

This is the version of AI-for-thinking I love. Not “AI does the thinking” — “AI makes the thinking I was already doing dramatically better.”

AI lets you create at the speed of thought

I run a raccoon mascot system on this blog. Cute anthropomorphic raccoon, rainbow round glasses, green t-shirt with bold white text, mismatched Crocs — sixteen of them now, one per major theme. Years ago, that was a project I’d have to commission. I’d find an illustrator, brief them, negotiate, wait, get something close-but-not-quite, iterate, pay again. Maybe it would happen. Probably it wouldn’t.

The current version of this project: I write a prompt, generate a batch, look at the choice sheet, pick the winner, ship it as a hero image on a post. The visual history walks through how it evolved across three generations of image models. The 7-habits marathon I did last spring needed seven on-brand raccoons — I generated four iterations of choice sheets (v1 → v4), Igor-the-human picked, and the post chapters got their visuals. That’s a creative loop a single person can now close in a weekend.

And it’s not just images. The hyper-personal thread doesn’t stop at words. Songs in your voice, a podcast where you interview your favorite dead philosopher, a bedtime story where your kid is the hero, a five-minute animated explainer of the boring concept you’re trying to teach. The medium isn’t the bottleneck anymore. Taste is the bottleneck. The skill is in knowing what’s good — what to pick from the choice sheet, what to throw away, what to push back on. AI multiplies whatever taste you bring. If you bring none, you get slop. If you bring some, you get more interesting work than you could have made any other way.

That’s the second thing I love. The creative ceiling for one person with taste just lifted by an order of magnitude.

Democratization of art

TODO: Alex — your take here on what AI is doing to the creator-class, the floor-vs-ceiling argument, and the gallery-system implications. Drop in 200-400 words.

Open questions for Alex:

  • Should this be your own guest post (full sub-post) rather than just a section embedded here?
  • What’s the definition of art?
  • Why was this kind of art-making impossible before AI? (Analogous to the strip-mall / entry-tax framing for Sergio’s school — what was the gatekeeping tax on art, and what specifically did AI collapse?)

AI clears the small-business tax — and the result is more diversity, not less

Back to Sergio.

The deepest thing the doomer narrative gets wrong is the homogenization argument. “AI will make everything the same — same content, same designs, same schools, all averaged out into beige.” That’s true if you assume AI replaces the human creator or operator. It’s the opposite of true if you assume AI gives the human their own staff.

Before AI: starting a small thing — a school, a practice, a label, a tiny shop with one specific aesthetic — required paying the entry tax. Compliance lawyers for the 501(c)(3) and the state filings. A curriculum consultant. A marketing firm for the prospectus and admissions materials. A development director to make the funding model work. Each of those professionals had their own clients to prioritize, their own rates that priced out anyone without capital, their own template solutions that homogenized the output anyway because they reused the same patterns across clients.

Most of that tax was the entry barrier. It wasn’t gatekeeping the quality of the schools — it was gatekeeping who got to start one. The result, in education: in Bremerton, Washington, and every American small town like it, the only options are the district public school, maybe one private school, and a couple of franchise tutoring chains in the strip mall. Three flavors of education for the whole town, because those were the only operators who could pay the tax.

AI doesn’t replace those professionals. It gives Sergio his own draft of each piece of their work so he can show up to the human conversations already 80% of the way there. The compliance lawyer still gets paid — for the hard call about which state requirement actually bites, not for the boilerplate. The curriculum consultant still earns their fee — on the deep pedagogy questions, not on aligning units to standards Sergio can now check himself. The marketing person still shapes the messaging Sergio can’t iterate to alone. The development director still works the donor network.

What changes is who gets to start. The STEM school Sergio specifically would have run, with his specific pedagogy, for the specific kids in his neighborhood — that school now exists. So does the seamstress’s shop, the bookstore the librarian always wanted, the small bakery that doesn’t fit a franchise template. Bremerton stops being one district school plus one private plus a tutoring franchise — it becomes a dozen small specific micro-schools, each one a real teacher with a real vision, because the entry tax collapsed.

And selfishly: every Sergio who ships their school is one less generic box store / box school in my world. Every weird specific shop with one person’s taste behind it is one less identical strip-mall tenant I have to walk past. The homogenization I hate gets thinner every time someone like Sergio actually clears the bar.

That’s not homogenization. That’s the opposite of homogenization. The doomer story has the polarity backwards.

Stuff I know nothing about, but I’m sure is going to be awesome: AI and medicine

I’m not a doctor. I’m not a biologist. I have no business having opinions on drug discovery or protein folding or radiology. But I watch the trajectory and I get genuinely excited — the rare-disease kid whose diagnosis used to take seven years and four specialists, the cancer treatment tuned to the specific mutation in your specific tumor, the imaging read that catches the thing the tired human eye missed at 2am. None of that is hypothetical anymore; it’s just early. And I don’t know the half of it, and that’s the point — the exciting stuff isn’t going to come from the people who already know the field. It’s going to come from outsiders who can finally stand on the shoulders of AI.

Change is hard, and previous winners will be losers

Yes, change is hard. Some of the people who built careers on the homogeneous, tax-paying world of yesterday — gatekeepers, professionals whose entire value was access to expertise that’s now democratized — are going to struggle in the new one. That’s true of every wave of progress: the industrial revolution, the internet, every step-change before this one ate the people who’d optimized for the prior plateau. Naming that honestly is part of being optimistic, not opposed to it. The transition will hurt some specific people in some specific ways. And the destination is still the right one — more agency for more people, more specific human stories shipped, more Sergios who get to start their school. But here’s why I’m still net-bullish on the destination, and why the doom case keeps under-shooting reality.

But what about the doom?

Every previous wave of pessimism has had its horse-poop moment. In the late 19th century, New York was projected to drown in horse manure within a few decades — the world’s first international urban-planning conference convened in NYC in 1898 specifically to solve the looming horse-shit crisis, and reportedly broke up early because nobody could see a way out (see Eric Morris, “From Horse Power to Horsepower”, Access Magazine, 2007). They were extrapolating linearly from current traffic onto current technology. Then the automobile showed up and the projected catastrophe became a non-issue overnight. That’s the move the pessimist always misses: today’s constraint gets dissolved by the next wave, not endured forever inside the current one.

Or take the “but we’re losing skills” panic. Sure — basically nobody can read a paper map anymore. And basically nobody is lost. And a lot of marriages got saved (the “left at the next light, NO the OTHER left” fight is mercifully extinct). Skills die, new layers emerge on top of them, and the net is more freedom and less friction, not less. The lost-skill is real; it’s just not the thing that matters.

Three objections deserve a sentence each.

“Jobs go away.” Some jobs do. New layers of agency open for people who couldn’t access them before. Net new first-time owners. The transition is real and painful; the destination is more individual agency, not less.

“AI homogenizes everything.” Only if you assume AI replaces the human. Sergio’s school is the counter-example. When AI is the staff and the human is the operator, you get more specific human stories shipped, not fewer.

“AI is slop.” Slop is bad prompting plus no taste. When you bring taste, AI multiplies it — see the raccoon project. The slop problem is a taste problem dressed up as a technology problem.

I’m not interested in litigating these at length. The point of this post is the optimism case, and the optimism case stands on its own evidence.