Build education machines.

I’m thinking a lot about AI + education these days.

When we think AI + education today the main things that come to mind are companies that make software that help students study (ex. upload a PDF get flashcards generated, upload a lecture get notes generated, etc).

Not much else comes to my mind at least.

There’s other stuff too. Like AI for teachers building lessons, or AI for people learning a new language, or AI tutors on top of existing video courses.

But, idk, all these just feel like…nothing that fresh.

AI is being used in the most wild, creative ways in 2025 — robotics, self-driving cars, solving cancer, drones, writing code. Every industry is being completely revolutionized in front of our eyes from manufacturing to software engineering to how we make memes.

Yet, when it comes to education we’re still making stuff that…

…helps students pass exams?

This is odd to me.

And it feels like we just aren’t thinking outside the box.

Learning, education, upskilling, studying — these words, they only describe the “act” versus the outcome. Very few people learn to just learn. They learn so they can do something.

A person learns sales so they can close their first deal. They learn video editing so they can start a YouTube channel. They learn design so they can go freelance and quit their job.

"Yes, Farza. This is obvious".

But…

If it's obvious…

Then why do I feel like we’re stuck in this loop with the education sector where education companies don’t understand how important they are? It seems like they are stuck building products that can get a person to pull out their wallet for "learning", and can be okay businesses, but, don’t actually change what a person can do in the world.

This leads to stuff like:

Language apps that give you a 500-day streak but you still can’t speak any real Spanish. Platforms with thousands of hours of content no one watches because its just boring. Certifications that look good on LinkedIn but don’t actually change what you’re capable of doing. AI tools that summarize lectures, generate quizzes, and help you cram for an exam you’ll forget in a week.

None of this is bad. But, it’s all optimized for the act of learning vs doing.

And, you can only monetize straight up “learning” so much.

How long or how much will someone really pay for something that just helps them learn without a clear result in sight?

Answer: not that much, and not that long.

This makes the market seems like an endless ocean of students that don’t wanna pay with terrible retention, jobless people without much to spend, or just random consumers who pay for a few years because they’re “lifelong learners”.

This leads education companies to seem “small”.

And, so investors don’t invest.

And, so founders don’t start education companies.

And, so we move on to other more lucrative things like verticalized B2B voice agents for customer service.

But, take a step back.

In reality, education companies are some of the most important companies to exist. They literally take a person, improve them, and create a version of them that is more capable.

In reality…

These aren’t even education companies. They are companies that build new human capital.

An education company that does it’s job literally creates new human capital in the form of inventors, designers, musicians, engineers, doctors, founders.

These are the people that change our reality.

When I look at Stanford, I don’t see a university.

I see a machine. You feed it an 18-year-old who’s a 3/10 in their ability to impact the world + pursue their goals, and four years later, a 6/10 on average walks out.

That’s pretty good.

That person who’s a 6/10 is now much more capable in doing things in which they can contribute back to the world.

And to me, we should be building more of these “machines” that build new human capital. Machines where a person who is a 2/10 in their ability to impact the world + pursue their goals in some domain walks in, and can walk out at least a 6/10.

Coursera, Udacity, or any of the course sites are interesting. Most people buy a course or two on there, never finish, and maybe go from a 2/10 to 3/10. Not terrible. But, meh. Not the best machines. (But, can be good businesses!). I feel similar about things like Duolingo, Skillshare, Masterclass

YC is one of these machines for later stage folks.

Most people walk in to YC as a 5 or 6/10 already, and walk out a 8/10 on average. Most YC founders would agree that the things that YC provides (money, mentorship, and customers) accelerates them in impacting the world + pursuing their goals.

Another machine that comes to mind is The Knowledge Society (TKS) which is one of my favorite education initiatives of all times. TKS is an after-school accelerator for high schoolers.

A 13-17 year-old comes in and TKS helps them build projects, gives pretty tight mentorship, and helps these kids go on to get internships, do breakthrough research, and starts companies.

An average 14 year-old high schooler can walk into TKS a 1/10 in their belief in themselves, their ambition, and their ability to work on hard problems — and if they can walk out even a 5/10 on average. That is a massive jump.

We thought about buildspace as a machine as well.

I figured if we create a machine where someone enters as a 2/10 and exits a 6/10 on average in their ability to start a business, reach their financial goals, and contribute to the world by working on their own ideas — then that system would be very valuable.

(Note: And it ended up being quite valuable. 1.5 years after the company closed, I still get messages near daily about how someone from a past season raised a round, got their dream job, or started a side hustle).

(Note: You can’t guarantee outcomes. Ex. “do this and you’ll make $100,000” or “do this and get a job at Google”. But, you can improve someones chances to reach certain outcomes)

So.

What am I getting at.

Well.

I think the biggest opportunity in AI + education isn’t in helping people study or learn random skills with AI.

It’s in building better end-to-end machines that output human capital. Machines that can take a person in random country at a 2/10 and actually get them to a 6/10 in their ability to pursue some goal with success.

Imagine different machines for:

A person who wants to become a YouTuber.
A person who wants to be an author.
A person who wants to learn to code and get a job.
A person who wants to get into sales.
A person who wants to become a filmmaker.

It’s going to be education companies that democratize the ability for people to actually become something.

And with AI, people have the tools to become anything. But, tools on their own won’t lead to mass change.

Think about what actually transforms people.

It’s not content. You can find content anywhere.

It’s offering the right balance of learning, accountability, and structure. All while properly maintaining your motivation.

The best human mentors do this naturally. They check in. They push. They know when you need encouragement vs. a kick. They care about whether you actually become the thing, not just whether you learned the thing.

One thing you’ll notice about machines that guarantee really good outcomes, is that they are simply difficult to scale.

YC supports around ~1,000 companies a year.
MIT accepts ~2,000 undergraduates a year.
TKS supports about ~1,000 kids a year.

These are all wonderful, wonderful, wonderful things. And I'm not trying to say "Oh only work on things that scale". I rather live in a world with MIT vs without.

Thing is, these things all have humans in their core loop.

That’s why the best “machines” — top bootcamps that guarantee jobs, top universities, elite apprenticeships — are so tough to access. They require humans to lead the initiatives.

And humans rarely scale.

So a tiny percent of humanity gets access to the best machines.

But for the first time ever, AI can change that.

I don't want this boil down too "Okay lets build AI Steve Jobs and we can scale mentorship to 100M people".

If that was the solution, it'd be done already.

So, think outside the box a bit with me for a sec.

I was in Pakistan a year ago — and I helped a 21 year-old there learn how to make a Shopify store, source clothes locally, and sell them on Facebook Marketplace. I had two check-in meetings a week with him.

And, we got him to making $250 a month! Which is how much his father made per month working as a driver.

This kid went from a 1/10 to 5/10 in front of my very eyes. He learned basic web design, he learned basic marketing, he learned customer service. And he took them all and built a basic business.

When I look back at those meetings, I can’t help but feel like I may as well have been an AI model he was talking to.

Of course, the human motivation element can’t be ignored, but, a lot of the stuff like: how I taught him Shopify, how I set his goals, how we adjusted the plan when things changed, how I helped him build a basic brand, how I helped him reframe his negative "I can't do this"…I saw all this and thought:

Could an AI have guided him to build that Shopify store?
Could it have taught him branding?
Could it have checked in with him everyday?
Could it have helped him set his goals?
Could it have told him where he was fucking up?

All that seems like stuff these models can totally be generalized for…

And hey if it works for a kid who wants build an e-commerce store, who's to say it wouldn't work for a kid in Morocco who wants to be a writer? Or a woman in Berlin who wants to be a designer? Or a person in India who wants to study genetics to one day solve epilepsy?

We’re not there yet but we’re getting there.

These models haven't been pushed to push humans yet. That's very under explored. But it's worthwhile. And we need really smart people to work on it.

The companies that figure this out won’t be “education companies.”

They’ll be human capital factories.

And they’ll be some of the most important, most valuable companies ever built.

Because at the end of the day, everything we care about — the startups, the cures, the art, the inventions — all of it comes from people. And the more capable people we create, the better our world gets.

P.S: I’m looking for a co-founder. I'm early and still not sure exactly what I'll be building. But, if this is something that seems like stuff you’d wanna work on lmk farza@humansongs.so or send it to someone that’d be interested.