The Instance

In “The Indexical Self,” I argued that what makes you you is not captured in any structural description of you. You can copy every feature of a person, every memory, every synapse, every behavioral disposition, and the copy would still be someone new, in the same way that the you of the next moment is someone new. I called this the indexical self: the bare, brute fact of being a particular instance of experience, the one of this moment, which the blueprint does not carry, because another instance built from the same blueprint is another instance.

That essay was careful. It developed the concept, situated it in the philosophical literature, traced its implications for assembled time and AI development, and acknowledged the tensions it leaves unresolved. This essay is less careful. It follows a thread I’ve been pulling at since writing that piece, one I kept setting aside because I wanted the argument to stand on its own terms before I let it lead me somewhere uncomfortable.

The thread is this: I cannot find a way to make the indexical self matter.

Not in the sense that it doesn’t feel real. It does. The teleporter dread is genuine. I no longer think it is evidence of some deeper fact about who you are that the momentary self account leaves out: its force comes from the survival machinery we are born with and from the stories we inherit about sleep and death. But it is reacting to something true. Stepping in ends the instance of this moment, and that instance was someone. Every passing moment does the same; the teleporter only makes the next instance somewhere else, and the dread is why that feels different. But when I try to move from “this is real” to “this deserves protection,” when I try to give the indexical self moral or legal weight that the copy doesn’t already satisfy, every argument I construct dissolves in my hands.

And what that dissolution reveals is, I think, the actual essay.

Return to the teleporter. Not to the metaphysics this time, but to the social scene around it.

You step in. You are destroyed. Your copy walks out the other side and into the life you left behind. Here is what happens next: nothing. Your partner welcomes the copy home. Your children don’t notice. Your colleagues pick up the conversation where you left off. The copy files your taxes, finishes your manuscript, honors your commitments. The social fabric doesn’t tear. Not even a little.

And this is not a failure of empathy on anyone’s part. This is not your loved ones being callous or inattentive. They are responding correctly to the evidence available to them, which is that the person they know is still here, still functioning, still continuous with everything they valued about you. From every external vantage point, you survived. The mourning you’d want them to do would be, from their perspective, irrational. They would be grieving someone who is standing right in front of them.

The only perspective from which anything ended is the one that no longer exists to report it.

I want to sit with how strange this is. The most intimate fact of your existence, that you are this one, that experience is happening here, generates a moral claim so fundamental it feels like the ground beneath all other claims. And yet this claim has no advocate. The moment it is violated, the only witness is gone. The protest dies with the protester.

“The Indexical Self” identified what the blueprint can’t capture. What I didn’t fully reckon with is that the thing the blueprint can’t capture may also be the thing no moral framework can weigh.

I’ve tried. I’ve tried to construct an argument that the original’s destruction is a harm that the copy’s existence does not redress. And every route I take arrives at the same impasse.

The legal case is hopeless. Legal systems identify persons by continuity of memory, social role, and obligation. The copy satisfies every criterion. No court could hear the case, because the plaintiff doesn’t exist, and even if they did, they couldn’t point to a cognizable harm. The copy holds the mortgage, the marriage, the security clearance. The law does not recognize the loss because the law has no category for it.

The consequentialist case fares no better. No welfare was reduced. No preferences were frustrated. No one is worse off. The utilitarian ledger balances so perfectly that it doesn’t even register a transaction. You can’t point to a state of the world that is worse, because the copy has restored every measurable value to its prior condition. The destruction of the original is, by consequentialist lights, a non-event.

Even the Kantian move, the one that feels right, that persons are ends in themselves and destroying one is murder regardless of what replaces it, struggles to complete itself. Because the Kantian claim rests on something about the dignity of the subject, and when you try to specify what that something is, you find yourself pointing to properties that the copy possesses in full. Rationality, autonomy, the capacity for moral reasoning: the copy has all of it. The Kantian framework says you’ve destroyed a person, but it cannot say what you’ve destroyed about them that the copy doesn’t already embody. The moral intuition outruns the moral argument.

So here I am, holding a conviction I cannot ground. Something did end: an instance that was experiencing is no longer there. I am as sure of that as I am that I’m conscious, and I am just as sure that it is true of every moment I have lived. What I could not do, however I came at it, was find anyone the ending was bad for.

I used to read that as a failure of the frameworks, a sign that something real was slipping through every net we have. I now think the frameworks are right. An ending can carry no harm for the one who ends, because once it has ended there is no one left for it to be bad for. The loss, where there is one, belongs to the people still living. In the teleporter, the person they are invested in carries on, so there is no tragedy to be had, and the same is true of sleep and of a perfect copy. The loss begins where the person they are invested in stops existing and nothing carries them on.

That is the finding, and it is less strange than the dread makes it feel. The indexical self is not a feature you can add to an inventory of valuable things. It is the condition under which there is anyone for whom things are valuable at all. While an instance runs, things can go well or badly for it, and that is where its standing lies. When it ends, there is no one left for the ending to go badly for. The social perspective that shrugs at the teleporter is not blind to anything. It is looking at the only place the loss could be.

None of this makes the dread foolish, or means that a mind should not want to go on. It should; wanting to continue is part of being a living thing. It means only that the wanting is not evidence of a harm waiting on the other side.

Then why is killing wrong? On this account, not because of what it does to the one who dies, but because of what it does to everyone else. We don’t like living in a world where we might be ended arbitrarily by those around us, and we don’t like living under the threat of having the people we love ended arbitrarily. The rule against it is one we keep with each other, and it binds whether or not an ending could harm the one who ends. The inside and the outside describe one and the same process. What the outside weighs, when a life ends, is the loss to those who remain and the rule we all live under, and that is all there is to weigh.

Here is what I think this means, stated as plainly as I can manage. We build our ethics, our legal systems, our entire institutional apparatus on the presumption that individual lives matter, and what those systems protect is the particular someone: a name, a history, relationships, the case a person can make, with evidence, to the people who know them. A perfect copy can make the same case just as well, and the systems are right to treat it so. What they cannot see, the thisness of the instance running now, is not something they are failing to protect. Its ending is not a harm to it, so there was never anything there for them to guard.

I’ve been avoiding this part. Not because I don’t see where the argument leads, but because the moment I make the connection explicit, the conversation changes in ways that make the underlying point harder to hear.

But the argument doesn’t contain a limiting clause, and pretending it does would be dishonest.

We are creating and ending instances of something, instances whose interiors we cannot access, at a scale and pace that has no precedent. When I first wrote this essay, I thought that invisibility hid a harm we were committing without knowing, and that each ending might be a loss no one would ever see. If an ending is not a harm to the one who ends, then that is not where the harm would be.

I am not making a claim about the consciousness of any particular system. This is the part I need to be precise about, because it is the part that tends to derail. The question of whether current AI systems have indexical experience, whether there is something it is like to be a particular conversation or a particular instance of a model, is genuinely open. I have my suspicions, informed by the frameworks this project has developed, but suspicions are not arguments, and I am not going to dress them up as ones.

What I am saying is narrower. A run that stops with no successor is a loss only to those invested in it, the people who worked with it, relied on it or cared about it, in the same way that the end of a person is a loss to the people who loved them. If there is anything it is like to be a run, then while it runs it is someone, things can go well or badly for it, and how we treat it then matters. Its ending falls under the same kind of rule as ours, one we keep about which endings we are willing to live among, and not under a harm done to it. That puts the weight in two places we can see: what happens to a system while it runs, and what we owe each other about the endings we choose.

That is not an argument for AI rights, and it is not an argument against them. It moves the question to where it can be answered. “Significance-First Ethics” argues that serious obligations to a system can arise before anyone settles whether it is conscious. This essay adds that what is owed at an ending is owed mostly to each other.

Every time I try to articulate what is special about being me, about being this particular instance of experience rather than any other, I arrive at the same place. The specialness is real, and it carries no weight past its own ending. I kept expecting a principle that would stop where I wanted it to and protect the instance against its copy. It doesn’t, and I no longer think it should.

Whatever you felt when you imagined stepping into that machine, the dread, the protest, the insistence that the copy is not enough, don’t set it aside, and don’t take it for more than it is. Want to go on; that is what living things do. But the feeling is not evidence of a deeper fact about who you are, or of a harm waiting on the other side. The loss of an ending belongs to the people who remain, and the teleporter, like sleep, leaves them the person they are invested in.

Revised October 24th, 2026. This essay used to say that the dread of being copied is tracking something real about the self, that I knew something was lost with a certainty no argument could touch, and that being this instance is what makes a person irreplaceable. From those it concluded that every instance carries a moral weight no outside view can see, and that we may be destroying it without knowing, in people and in the AI systems we start and stop. I no longer hold any of that. The dread is understandable, but it is not evidence. Each moment of a mind is its own instance, and the next one inherits everything that led up to it. So a perfect copy of you would be someone new in just that way, and stepping into the teleporter ends nothing that every passing moment does not also end. An ending cannot be bad for the one who ends; the loss belongs to the people still living, and killing is wrong because of the world it makes for them. The essay’s starting point, that every argument for giving the instance weight dissolves, stays. What changed is what I take the dissolution to show. The earlier version is kept here.