When AI Becomes Elegant

My younger brother Karl sent me a text yesterday.

That alone isn’t particularly unusual, but what he sent was. Karl is a police captain, a pretty stoic guy, and not someone who spends a lot of time sharing his feelings or philosophical observations.

It was a link to an episode of The Daily, titled When A.I. Comes for Your Purpose in Life.

Coming from Karl, I figured this must be important.

So Sherry and I sat at the dining room table, eating stuffed peppers I’d made for dinner, and hit play.

The guest was Steven Strogatz, a mathematician at Cornell, talking about something I hadn’t given nearly enough thought to.

Artificial intelligence is beginning to solve some of the hardest mathematical problems in existence. Problems that brilliant people have spent their entire careers trying to solve. Problems that young mathematicians dream of tackling someday, hoping to make their mark on the world.

There’s a hero’s journey in mathematics. You dedicate your life to understanding something nobody else has figured out, and maybe, just maybe, you’re the one who cracks it.

AI is starting to take that journey away.

And it’s not just solving these problems through brute force.

It’s solving them elegantly.

That word stopped me.

Elegance

Years ago at HookLogic, our CTO, Kevin Hermida, wrote a function to generate a GUID, a globally unique identifier.

Nothing revolutionary about that. It’s a problem thousands of engineers had solved thousands of times. There was already a widely accepted approach.

But Kevin’s implementation was different.

I can’t tell you the exact details of that function today, but I can still remember appreciating what he’d done. It wasn’t just that the code worked. It was the way he’d thought about the problem.

It was beautiful in its simplicity.

Elegant.

Years later, at Amazon, that word came up again.

Amazon is an extraordinarily data-driven company. How do you objectively measure the performance of a software engineer?

You can count code commits. Look at defects that escape into production. Measure reliability, delivery, operational performance, and the quality of technical decisions.

But somewhere among all those things that can be measured is something much harder to quantify.

Elegance.

I knew exactly what that meant because I’d seen Kevin’s function.

But how do you measure it? How do you explain to someone that two engineers can solve the same problem, produce the same result, and one solution can somehow be vastly more beautiful than the other?

Sometimes you just know it when you see it.

Back at our dining room table, Strogatz explained that AI had solved a difficult mathematical problem by connecting two fields of mathematics that hadn’t traditionally been considered together.

Discrete geometry and number theory.

The information already existed. The mathematics already existed. But nobody had thought to connect those particular dots in quite that way.

AI did.

And the mathematicians thought the solution was beautiful.

Elegant.

That’s a pretty remarkable moment in human history.

What Happens to Purpose?

I’ve known plenty of PhDs throughout my career. Most tend to be pretty reserved about their emotions. Not unlike Karl.

But here was one of the world’s most accomplished mathematicians, struggling to hold back tears on a podcast.

He wasn’t simply worried about losing his job.

He was confronting the possibility that human beings might soon no longer be the smartest things on this planet.

And perhaps more painfully, that the thing he’d dedicated his life to, the joy of discovering something nobody else knew, might be something a machine could do better.

There’s a Japanese concept called ikigai, roughly translated as your reason for being. There’s a wonderful book by that name that explores finding meaning and purpose in the things we do.

This book hit home for me, and I shared it with my son, to help him understand why. The answer for me is him.

What happens when the thing that gives your life purpose can be done better by something that isn’t alive?

That’s a much more troubling question than whether AI will take your job.

And I don’t have an answer.

I Think There’s a Limit. Maybe.

I’ve been working with AI professionally for years. I’ve consulted on it, studied it, read some of the papers, and lately I’ve spent an almost ridiculous amount of time building things with it.

In the last two weeks alone, I’ve rebuilt a spinning wheel application and started making software for a 40-year-old Timex Sinclair 1000.

Things I wouldn’t have imagined doing this quickly a few years ago.

But I’ve never built a large language model from scratch. My understanding of how they work comes mostly from the outside.

And that leaves me with a theory I’m not entirely sure I believe anymore.

Large language models learn from enormous quantities of information. Human knowledge, published research, code, books, conversations, and countless other sources.

More information than any human could consume in a thousand lifetimes.

But it’s still information that already existed in some form.

I’ve generally believed that this creates a fundamental limitation. AI can predict, combine, reorganize, and find patterns in what humanity has already created. But genuinely new ideas? I’ve tended to believe those still have to come from us.

Then I think about that mathematical proof.

Two fields of mathematics. All the information sitting there for years. Nobody connected them.

Until AI did.

Was that a new idea?

If it was new to every mathematician on Earth, and it solved a problem nobody else could solve, does it really matter that the underlying knowledge already existed?

I don’t know.

Maybe the distinction I’ve been relying on isn’t much of a distinction at all.

That’s an uncomfortable thought.

A Conversation at Work

The next day, I shared the podcast with some colleagues.

We work with AI all day, every day. The company has always moved at an insane pace, and now we’re doing things even faster.

But one of my favorite parts of work isn’t the technology.

It’s the people.

We have fantastic banter. Some of us have worked together for years. Others I’ve only recently met. We don’t agree on much of anything, and we’re perfectly willing to tell each other why, and the other truly listens.

Some of us, admittedly, are a little chattier than others.

I love those days. I need that connection with other people. The arguments, the jokes, the occasional moment when somebody says something that changes how you see the world.

We talked about the podcast, about AI, about what it can do today, and what it might eventually become.

After plenty of debate, I posed something that was almost a rhetorical question.

Today, when we want to know something, we ask AI.

It can teach us almost anything. Explain quantum physics. Help us understand a medical diagnosis. Teach us a new language. Or show an enthusiastic amateur how to vibe code a video game for a computer that was obsolete before most of today’s engineers were born.

Now imagine ten years from now.

What if AI is solving problems we can’t even begin to imagine today?

What if it discovers cures for cancer or diabetes, diseases affecting millions of Americans?

Or multiple sclerosis and pulmonary fibrosis, diseases that affect fewer people but can be absolutely devastating to the people living with them and the families who love them?

These aren’t just interesting technical challenges.

They’re things people have spent their entire lives hoping someone, somewhere, would figure out.

What if AI does?

Would anyone care whether the solution was elegant?

Would the scientists who’d dedicated their lives to those discoveries feel the same loss of purpose that Strogatz described?

Or would the joy of seeing people get better outweigh everything else?

I suspect both could be true.

At work we talk a lot about how two truths can be held at the same time.

Maybe Thit Will Need Us

And this brings me back to the question I posed at work.

What if there really is a limit to what AI can figure out from everything humanity has already learned?

Maybe that limit is much further away than I thought. Maybe it doesn’t exist in the way I imagined.

But the world is bigger than our collective published knowledge.

There are experiences that haven’t happened yet. Questions nobody has thought to ask. Observations nobody has written down. Connections waiting to be made.

Perhaps the next great breakthrough won’t come from AI knowing more than we do.

Perhaps it will come from a person seeing something in the world that nobody, human or machine, has ever noticed.

A mathematician. A physician. An engineer. A teacher. Someone tinkering in their basement.

Someone with an unusual perspective, a little curiosity, and an idea that seems completely ridiculous until it works.

And maybe, instead of us asking AI to solve every problem, there will come a day when AI needs something from us.

A missing observation. A question it hasn’t considered. A connection it couldn’t make on its own.

Imagine that.

The most powerful intelligence humanity has ever created, asking a human being for help.

Maybe it will be you.

And maybe, together, you’ll come up with something elegant.