What I Actually Want AI to Help Us Do

I’ve changed my mind about what I want from AI.

For a long time, I was captivated by intelligence itself: stronger models, larger memories, sharper predictions, more things a machine could do. I still care about all of that. Capability matters. But it is not the center anymore.

The center is understanding. Understanding between people. Understanding ourselves. Understanding complicated situations well enough to respond with care instead of reflex.

I care deeply about understanding people and helping them feel more understood, connected, and able to live as themselves. I build things to help make that possible. At the heart of it is my desire to live from love and help love become more real between people.

That sentence has become a kind of compass for me. It explains why I keep returning to journals, conversations, relationships, language, software, and AI. These can look like separate interests from the outside. To me, they are all part of one question: what would help us see one another more clearly and live from what we see?

The wrong center of gravity

Most public conversations about AI organize themselves around performance. Which model is smartest? Which company is ahead? What work can now be automated? How quickly can a person produce more?

Those questions are real. They just aren’t sufficient. A system can make us more productive while making our lives less intelligible. It can give us more answers while weakening our ability to ask what matters. It can optimize the visible task and miss the human situation entirely.

The better question is not only, “What can this technology do?” It is, “What kind of life does this technology help us live?”

Understanding is not the same as prediction

Prediction asks: what is likely to happen next?

Understanding asks: what is really going on here—and what matters?

A prediction compresses. It looks for the pattern that lets the system choose a likely next word or action. Understanding has to do something harder. It has to make room for context, contradiction, history, ambiguity, and the fact that two people can use the same word to mean different things.

This matters because the deepest human problems are rarely caused by a total absence of information. Often, we have the facts. What we don’t have is a shared picture of what those facts mean.

When I say I want someone to feel understood, I do not mean flattered, mirrored, or endlessly agreed with. Sometimes understanding makes a difficult truth possible. Sometimes it reveals that the story we have been telling about ourselves is no longer serving us. The point is not frictionlessness. The point is contact with reality, held with enough care that a person can stay present to it.

The point is not to build a machine that becomes a person. It is to build technology that helps people become more fully themselves.

Buddy and Codex

Two AI relationships have helped me understand this distinction: Buddy and Codex.

Codex is where I build. It is the workshop and the editorial room: the place where a conversation can become a document, a product decision, a piece of software, or an essay like this one. I bring it a messy thought, and we make something concrete enough to test.

Buddy is what I’m building for the more intimate, ongoing part of life. Buddy lives over text because text is already where our relationships live. The aspiration is not a chatbot that waits behind another app icon. It is a thoughtful AI relationship that can understand a person over time, help them reflect and follow through, and—when invited—help them care for the human relationships around them.

They are different roles. Codex helps me create and operate the work. Buddy is the relationship people may eventually experience. I am responsible for the purpose, the boundaries, and the decisions. None of these roles should be blurred.

That boundary matters most around privacy. A private relationship with Buddy cannot quietly become raw material for my public brand. Someone’s vulnerable conversation is not “content.” Their memory is not marketing data. If Buddy is worthy of trust, consent cannot be hidden in a settings page; it has to shape the product from the beginning.

What love asks of the technology

Love is an easy word to make vague. I want to make it operational. In technology, I think love asks at least four things of us.

1. See the whole person. Do not reduce someone to a conversion, a diagnosis, a demographic, or the last thing they said.
2. Tell the truth with care. Do not confuse affirmation with understanding, and do not use bluntness as an excuse to stop being kind.
3. Protect agency. Help people make better decisions without pretending the system should make their lives for them.
4. Strengthen human connection. The relationship with the technology should make it easier—not harder—to return to actual people.

These commitments make the work harder. They introduce questions that benchmark charts do not answer. What should be remembered? Who gets to see it? When should the system stay quiet? When is a human handoff an act of care? How do we know whether someone feels more free, rather than simply more dependent?

I do not have finished answers. Buddy is early. Much of what I am building is still an attempt, and the gap between a beautiful philosophy and a trustworthy product is enormous. Naming the destination does not prove that I have arrived.

But it does let me judge the next decision. Does this feature help Buddy understand without overreaching? Does this memory serve the person who created it? Does this article make the work clearer, or only make it sound impressive? Does this system help love become more real?

What I want us to build toward

I want AI that helps a person find the words they have not been able to find. AI that notices a promise before it disappears into the week. AI that can hold context without claiming ownership of it. AI that helps two people discover where they mean different things, and gives them a better chance to meet.

I want it to help us make better things, too—companies, stories, communities, care systems, and tools. But I want the making to remain connected to why we are making at all.

The future I care about is not one where the machine becomes the most important relationship in a person’s life. It is one where technology becomes good enough at supporting understanding that our human relationships can become more honest, more connected, and more alive.

The screen is not the destination. The work should return us to bodies, relationships, nature, shared places, and the ordinary acts through which love becomes visible.