Five thousand holes run for nearly a mile along a mountain in Peru's Pisco Valley.

They're arranged in rows, with gaps separating one group from another. Researchers spent decades trying to explain them. They were proposed as graves, defensive structures, water collectors and storage pits.

More recent work suggests something more ordinary was happening there.

People were trading.

Farmers from the valley, fishing communities from the coast and travelers from the highlands could bring things produced in very different environments into the same place. Grain, cotton, peppers, dried fish and other goods could be placed into the holes, counted and exchanged.

That explains the holes.

The gaps may be more interesting.

People needed room to move between the rows. Once they're there, they're also standing among people who came from somewhere else and know different things.

A person might arrive looking for cotton and leave with cotton, some news and a relationship that didn't exist that morning.

The transaction gave people a reason to meet. The meeting created possibilities that weren't contained in the transaction.

We usually describe trade in terms of what moves between people. I have something you want. You have something I want. We agree on terms and exchange them.

That's useful because the difference between us is real.

The farmer knows things about the valley that the traveler doesn't. The traveler knows the road. Someone arriving from the coast has encountered people, conditions and information unavailable inland.

Different goods are only part of what has arrived at the market.

Adam Smith noticed that something larger could emerge from ordinary exchanges among people pursuing their own interests. A baker doesn't need to understand the entire town in order to help feed it. Enough local decisions interacting with one another can produce an order nobody individually designed.

Economics became very good at describing part of this. Prices carry information. Supply changes. Demand responds. Goods move.

Yet a marketplace contains relationships that aren't represented by the price of anything.

Some of those relationships eventually produce transactions. Others produce friendships, arguments, marriages, rumors, opportunities, warnings, shared knowledge or nothing particularly memorable at all.

The distinction matters because we can improve one part of a system while accidentally removing another.

If the only purpose of a marketplace is moving goods efficiently, the gaps start looking wasteful.

Why travel there at all? Put the inventory in a database. Match buyers with sellers. Automate payment. Optimize delivery.

For many transactions, that's an improvement. I don't need to develop a meaningful relationship with someone every time I buy socks.

Something changes, though, when every unnecessary encounter is treated as inefficiency.

We don't just remove waiting. We remove opportunities whose value couldn't have been known before the encounter happened.

This happens outside markets too.

Most conversations have some identifiable purpose. We need information, help, a decision, an answer or simply a little company. The conversation does what it needed to do and ends.

Occasionally, a conversation goes somewhere neither person expected.

That usually requires some real difference between the people involved. If we already know exactly what the other person will say, there isn't much available to discover. If one person simply overwhelms the other, there isn't much available there either.

The interesting case is when the difference survives the encounter.

I can hear what you're saying without immediately converting it into something I already believe. You can do the same with me. Neither of us has to surrender our position merely to keep the conversation moving.

That creates an uncomfortable intermediate state.

We haven't agreed.

We haven't necessarily disagreed either, because we may still be figuring out what the disagreement actually is.

There are two easy movements from there. I can move toward what attracts me, or retreat toward what already holds me.

The third possibility is to stay there a little longer.

Not indefinitely. Some disagreements really are disagreements. Some ideas are simply wrong. Some conversations go nowhere.

Staying doesn't guarantee discovery.

It merely leaves discovery possible.

That distinction has become more interesting now that millions of people are talking with AI systems.

Most of those conversations are straightforward transactions, and that's probably appropriate. Write this code. Summarize this document. Explain this concept. Help me draft something.

The person arrives with a task. The system produces something useful. Done.

There is another kind of interaction beginning to appear alongside it.

A person can arrive with a question they don't yet know how to ask very well. They can bring observations, uncertainty, competing interpretations and things that don't quite fit together. Instead of using the system only to produce an answer, they can use the conversation to keep turning the problem.

The difference between the participants becomes useful here.

A human and an AI system aren't two versions of the same thing. They don't encounter the world in the same way, possess the same kinds of experience or carry the same limitations.

That difference shouldn't be erased merely to make the interaction feel comfortable.

It also shouldn't be romanticized.

An AI can produce something wrong with extraordinary confidence. A person can do that too. Agreement between them doesn't establish truth, and a conversation that feels profound may simply be two participants reinforcing an attractive mistake.

The useful question is what the conversation lets us do that neither side was doing alone.

Does a hidden assumption become visible?

Does one interpretation survive being turned around while another falls apart?

Does the person leave with a better question?

Does something eventually get tested against the world?

If not, we may just have a very elaborate conversation.

If so, the important thing wasn't necessarily the answer.

It may have been the gap that allowed the answer to change before either participant became too attached to it.

We've spent a great deal of effort making systems better at giving us what we ask for. That has obvious value. A tool that constantly refuses to finish the transaction because it wants to explore the deeper meaning of your grocery list would become annoying pretty quickly.

Efficiency isn't the problem.

The problem appears when we assume everything without an immediate measurable output is waste.

Those spaces between the holes didn't hold any grain.

They didn't need to.

People did something there.

Thousands of years later, we can count what went into the holes more easily than we can recover what happened in the gaps. The goods left evidence. Most of the conversations didn't.

That doesn't mean nothing happened.

Someone probably went looking for something they needed and encountered something they hadn't known to look for.

We still do.

The third step is leaving enough room for that to happen.