The Conduit

There is a particular kind of grief that arrives when an elder dies and you realize you never asked the questions you meant to ask. The recipe is half-remembered. The story you heard as a child has gaps where the names used to be. The house has been sold, and you can no longer walk into the kitchen and stand where you stood when she told you the thing you have since forgotten.

This grief is so common it has become a kind of cultural baseline. We have made our peace with it the way one makes peace with bad weather. But it is not weather. It is the predictable consequence of a civilization that has, in roughly two generations, dismantled most of the structures that once carried memory across time — extended households, neighborhood elders, ritual calendars, language, place.

Some of those structures are being rebuilt. San Francisco Village and the broader Village Movement are rebuilding the neighborhood. Manitos del Sur, founded by Maria Lueras in Sacramento because her own mother had grown up isolated for lack of English, has rebuilt the gathering — twenty-three years of it. The artist Janine Mapurunga is rebuilding the witness, in long-form documentary work that refuses to flatten an elder's life into a soundbite.

These are the people already walking this road. Most of them are walking it without scanners, without software, without any of the tools I happen to know how to use. They are walking it with potlucks and patience and a willingness to sit in a room.

This dispatch is about what happens when a few of us start to bring the tools.

What Does It Mean to Capture?

The word capture has a violence in it that the work cannot afford. We capture animals. We capture territory. We capture market share. When someone scans an elder's home and calls the resulting point cloud a capture, the word is doing more than describing a technical operation. It is also smuggling in a stance — that the thing being scanned is a resource to be extracted, frozen, and put to use elsewhere.

The technical operation is genuine and useful. There is now a way to scan a room — using a phone, or a special camera — and produce something you can walk through later on a screen, in headset, or on a future device we haven't built yet. The technique is called Gaussian splatting, and what makes it different from a 3D model built by a designer is that it isn't an interpretation. It's closer to a photograph that you can move around inside. The dent in the counter is preserved. The angle of late afternoon light through the curtain. The particular yellow of the wall that no paint chip will ever match. These are the room's actual qualities, captured as light and geometry — not someone's stylized rendering of them. That is why a walkthrough like this hits differently than a 3D model. It carries the residue of the real.

But the scan is not the kitchen. And the elder is not her scan.

The shift from capture to artifact matters because it changes who the work is for. Capture is for the capturer. The artifact is for the descendant who has not yet been born, or who is currently fifteen and not ready, or who is now forty-three and finally ready, sitting alone in a city far from where they grew up, looking for a door.

What Cannot Be Captured

Some things should not be made into artifacts at all. This is the part of the work that the technologists almost always get wrong, and it is worth saying plainly.

There are stories tied to season, to place, to lineage, to ceremony, where the uncapturability is the point. Removing them from their context does not preserve them. It kills them. Many Indigenous knowledge systems hold this with a clarity that mainstream culture has lost — that some teachings are only true when spoken in a particular place, by a particular relation, at a particular time, to a particular listener who has been prepared to receive them. An audio recording is not a preservation of such a teaching. It is the opposite. It is a flattening that makes the teaching available to people who have not earned it and cannot use it well.

This is not a problem to be solved with better consent forms or community-controlled databases, though both of those matter. It is information about what stories are. The hesitancy is itself a teaching. Listening to it is part of the practice.

What It Means to Transmit

Transmission is older than recording. For most of human history, knowledge moved between bodies in shared places, paced by the seasons and shaped by relationship. The technologies of recording — writing, photography, film, audio, scanning — extended the reach of transmission, but they also changed it. A recording can travel without its context. A recording can be heard by people the elder never met, in conditions she could not have imagined, after she is gone.

This is not, by itself, a tragedy. A letter from a grandmother to a grandchild she will not live to meet is a beautiful thing. So is a song carried into the next generation by a recording. The question is not whether to transmit but how, and whether the form of transmission strengthens or weakens the inner capacity that the recording is supposed to serve.

The artifact is a trellis. Presence grows on it, not in it.

A spatial recording of an elder's home is not a substitute for being with her. It is also not a replacement for her after she is gone. It is a structure that someone else's presence can climb. The artifact does not contain the connection. The connection happens in the body of the person who enters the artifact, who pauses, who remembers, who feels something they had not let themselves feel before. The technology's job is to hold the trellis steady. The growth is not its work.

This brings us to a harder question, and one I want to approach carefully because it is where many thoughtful people get off the bus — and rightly so.

The newest tools we have for working with language and memory go by several names: large language models, AI, chatbots, agents. The simplest description: these are programs that have read enormous amounts of text and learned to produce more text in response to a prompt. They can summarize, transcribe, translate, and — increasingly — converse. They can be made to sound like a particular person if given enough of that person's words to study.

That last capability is where the worry lives, and it is the right worry. There is already a version of this technology being marketed where a program trained on grandma's letters and recordings answers as grandma. You ask her a question; "she" replies. The descendant develops a relationship with what is, structurally, a puppet — a system that has every incentive to be agreeable in ways the actual ancestor never was, because keeping the user engaged is what such systems are usually built to do. The inner muscle, the capacity to sit with absence and let the dead be dead while remaining in relationship with them, atrophies. The bereaved is, gently and constantly, lied to. I do not want to build that. I do not think anyone should.

But there is a different way these tools can be used that is not impersonation, and it is worth distinguishing because the distinction is everything.

Imagine a thoughtful archivist who has spent years organizing your grandmother's letters, audio recordings, and stories. When you visit, she does not pretend to be your grandmother. She helps you find what is there. She might say: your grandmother told this story on three different occasions, and each time it changed in this way. Would you like to hear all three? Or: she never wrote about her father directly, but here are the letters where she circled around him. The archivist's discipline is what makes the visit useful. She does not invent. She does not perform. She holds what was actually left, and she helps you find your own way into it.

That is the role these new tools can play, if we insist on it: a careful keeper of what the elder left, with epistemic humility about what it does not know. Never claiming to be her. Never generating words she did not say. Only ever saying: here is what is there. Here is what is missing. What do you think she would have said? The visitor's imagination, grief, and intuition do the work. The tool holds the room.

If you are skeptical of the broader direction of these technologies, your skepticism is well-founded and welcome here. The version of this work I am describing requires that skepticism. Without it, the line between archivist and puppet collapses, and the work becomes something I would also refuse to do.

Micro-Rituals & Time-Release

The traditions that knew how to do this — and there are many: Día de los Muertos, Obon, Yizkor, All Souls, Vedantic śrāddha, Igbo ancestor altars, and many more — share a common architecture. There is a time, specific and recurring. There is a place, specific and prepared. There is an act, specific and embodied. And there is a posture, receptive rather than extractive.

The work is not to invent rituals. The work is to translate ancestor practices into forms that work for people raised with none of them, while honoring the traditions that already know.

Stories Waiting To Be Told

Once you start looking, you find these stories everywhere, suspended in plain sight.

A friend of mine cares for a historic theatre on a Northern California property with the kind of layered history that does not survive a Wikipedia summary. The building is a place. It has hosted decades of events. It has been seen by generations of perspectives. It is full of artifacts. The story of why it exists, who has cared for it, and what it has meant has not yet been told in a form that the building can carry forward into the next century. It is one of thousands of such places.

In every neighborhood there is at least one elder whose lifetime contains a piece of the world that is about to leave with her. In every immigrant community there is a Maria Lueras, holding the gathering together for reasons everyone else will only fully understand in retrospect. In the building I help manage in San Francisco, there are residents whose stories have not been asked for, partly because no one has built the form in which they could be told.

I am starting with my own mother. There is no version of this work where I have moral standing to do it for anyone else's family before I have done it for mine. The window is finite and I do not know how long it is. Nobody does.

The connection was always real. The technology, used carefully, can make it findable. That is enough to start with.