Feature proposal: Find someone you can think with endlessly
I’ve been thinking about something that AI might eventually be uniquely capable of doing.
We already use AI to find information, explore ideas, solve problems, and sometimes simply to think out loud. But what if AI could also help us find each other?
There are billions of people in the world. Somewhere among them may be a person with whom you could think for hours — not because you agree on everything or have identical interests, but because every idea from one person creates another idea in the other.
The problem is that you may never meet.
So I would like to propose an optional ChatGPT feature with a simple goal:
Find someone I can think with endlessly.
What I mean by that
Most matching systems are built around similarity.
Social networks know what content we like. Dating apps know our preferences. Professional networks know our skills and industries. Recommendation algorithms can find people who resemble us in hundreds of different ways.
But similarity is not necessarily what makes a conversation intellectually interesting.
Two people can share almost all their interests and quickly run out of things to say. Two completely different people can begin discussing one small question and, hours later, still be generating new questions from it.
Imagine:
A proposes an idea.
B notices an assumption A missed.
That changes the idea for A, who develops it in another direction.
B connects the new version with something neither of them was originally discussing.
That connection changes the original question again.
Eventually they arrive somewhere neither of them was likely to reach alone.
That is the thing I would want the system to search for.
Not agreement. Not similarity.
Continuation.
Why ChatGPT could potentially recognize this
Something interesting happens in long conversations with AI.
On social networks we usually publish the result of our thinking. In conversations with ChatGPT, we often expose the process itself.
We say things like:
“What if…?”
“No, that doesn’t work.”
“Wait, go back.”
“What happens if we reverse the assumption?”
“I don’t know why, but I think these things might be connected.”
Over time, these conversations can reveal patterns: how someone forms hypotheses, reacts to contradictions, explores alternatives, connects distant concepts, abandons ideas and later returns to them.
With explicit permission, perhaps ChatGPT could model not only what a person thinks, but something about how their thinking moves.
Then, among other users who have also explicitly chosen to participate, it could look for complementary reasoning patterns.
The question would no longer be:
How similar are these two people?
It would become:
What happens when these two people think together?
Thought Continuation
We could call this property Thought Continuation.
A possible Thought Continuation Score would not measure how long two people can keep talking. A conversation can be very long and still go in circles.
Instead, it could try to estimate whether interaction repeatedly creates meaningful new directions: productive disagreement, new conceptual branches, hidden assumptions being discovered, connections between distant subjects, transformations of each other’s ideas, or conclusions neither person was likely to reach independently.
In other words, the best match might not be the person who is most similar to you.
It might be the person who repeatedly makes you think:
“I would never have gone there on my own.”
How it could work
The feature should be completely voluntary.
A user could choose something like:
“Allow ChatGPT to look for people whose thinking may interact unusually well with mine.”
Only users who explicitly opt in would enter the matching pool.
Their original ChatGPT conversations would remain private. ChatGPT could instead create an internal reasoning profile specifically for matching.
If the system found an unusually promising match, both users could independently receive a message:
There is someone whose way of thinking may be unusually complementary to yours. Would you like to start an anonymous conversation?
Nothing would happen unless both people agreed.
The first conversation could even begin without names, photos, professions, countries or other personal information.
Just two people and a question.
If they later wanted to reveal more about themselves, that would be their choice.
Privacy
For something this personal, privacy could not be an afterthought.
Participation should require explicit opt-in. Conversations should never be exposed to another user. Derived reasoning profiles should not be public. Contact should require mutual consent. Users should be able to leave the matching system and delete their matching profile whenever they want.
And the purpose should be to understand how two reasoning processes interact, not to secretly classify people using sensitive personal characteristics.
Why I think this could matter
AI has become remarkably good at connecting a person with information:
person → AI → information
But perhaps there is another possibility:
person → AI → person
For most of human history, the people we could think with were largely determined by geography, language, school, profession, social circles and chance.
The internet put billions of people within theoretical reach of one another.
But it did not solve the problem of finding the particular person whose mind makes yours go somewhere new.
Maybe AI can.
I don’t want an algorithm to find someone who simply agrees with me.
I don’t even necessarily want it to find someone like me.
I want it to find someone I can think with endlessly.
Because somewhere among millions of people, there may already be two minds capable of an extraordinary conversation.
They just haven’t found each other yet.