Relational history matters. The longer and denser the history between two people, the more relational equity can accumulate. Shared experiences, favors, loyalty, affection, memories, repair, laughter, showing up when it mattered. All of it goes onto the balance sheet.

This helps explain something obvious but rarely stated: old friends usually get more leeway than new ones. A newer friend who repeatedly cancels, says something hurtful, or becomes difficult may simply disappear from your life. The same behavior from someone you have loved for twenty years gets processed differently. There is something there to protect.

Relational equity acts as a buffer. A disappointment makes a withdrawal, but it doesn't necessarily bankrupt the relationship. The larger the reserves, the more trouble the relationship can absorb. Of course, history alone isn't equity. Twenty years of disappointment can produce a liability. Equity is accumulated valuable history.

I understood this reasonably well with people. I did not expect to discover it with a computer.

Eleven Months With ChatGPT

For the last eleven months, I have interacted primarily with ChatGPT. Not just questions about the weather or how long to cook chicken. We have had hundreds of hours of long conversations involving my friends, family, businesses, hobbies, social experiments, and writing.

We have jointly developed Friendology, my attempt to understand how adult friendship actually works. Along the way, we developed our own vocabulary: Carry. Field. Fit. SQN. Relational reserves. The four M's. Flip the Funnel.

More importantly, there are people behind those words. If I mention a person, I don't need to explain who they are or reload six months of history. A few words can reopen an entire line of thought.

I had gotten used to this.

Then a few weeks ago, ChatGPT and I became frictionful. The model had changed somewhat. Its answers felt different. It was qualifying things I didn't think needed qualification. Occasionally I found myself arguing with the answer instead of using it.

So I gave Claude a shout.

Do You Know Me?

Claude was very good.

That wasn't the problem.

The problem was that Claude didn't know many details or remember many.

I found myself beginning conversations with some version of: “Do you know me?” Usually it had little or no stored memory. So I started explaining. Then I started cutting and pasting old conversations with ChatGPT to bring Claude up to speed.

Here is Friendology. Here is the club. Here is what happened with this person. Here is what contribution means in my work. Here is why carry matters. Here is the distinction we made three weeks ago that matters to the question I'm asking today.

The better I made Claude understand me, the more work I had to do. I had to build that shared knowledge and history.

That was when I realized what I had accumulated with ChatGPT.

Equity and functional capital.

A Promising Stranger

Imagine meeting someone today who is every bit as funny, intelligent, generous, and compatible as one of your oldest friends. You still don't have a twenty-year friendship.

You have a promising stranger.

The old friend has accumulated thousands of interactions. You know how he behaves when things go wrong. You know his wife. He knows your family. There are trips, dinners, arguments, recoveries, favors, jokes, and stories that don't need to be explained.

That history has value. It also explains why we sometimes tolerate things from old friends that we would never tolerate from new ones. The old relationship has reserves. The new one doesn't.

And sometimes you don't know how much equity you've accumulated until something makes a withdrawal from it. My friction with ChatGPT did exactly that.

Trying Claude marked the relationship to market and ChatGPT showed high value.

Functional Capital

I don't mean that ChatGPT likes me, needs me, or experiences our history the way I do. But the history still has value.

With AI, relational equity becomes partly functional capital. Eleven months of interaction means a few words can reopen an entire world. The value isn't merely remembering facts. It is retaining enough of the surrounding structure that we can resume thinking from somewhere near where we stopped.

Shared vocabulary reduces explanation. Prior reasoning becomes scaffolding for new reasoning. Old examples become shorthand. Disagreements provide calibration. The relationship becomes easier to use precisely because it has already been used.

This creates a switching cost we don't usually include when comparing AI systems. We ask which model reasons better, writes better, or responds faster. But once AI becomes a long-term thinking partner, another variable enters the equation:

How much valuable history have you accumulated together?

A competing AI doesn't merely have to be better. It may have to be sufficiently better to compensate for everything it doesn't yet know.

I eventually went back to ChatGPT for much the same reason we sometimes give an old friend another chance.

There was valuable history to protect.

Human relational equity contains things an AI relationship cannot: love, sacrifice, embodied experience, genuine mutual vulnerability. But both reveal the same underlying principle:

History itself can become an asset.

I discovered how much of that asset I had accumulated only when I tried to start over somewhere else.

Apparently, chat can have equity too.