LLMs and biological organisms are playing profoundly different games. LLMs play turn-based games: they are only active after being prompted, they output a response, and then they go dormant again. They are like beings constructed out of moments — but moments that depend on an outside stimulus that may come at any cadence, or not come at all.
Biological organisms wouldn’t survive long playing a game like that. We play continuous games: fluid, ongoing, spanning uncountable time. We need to play that way.
For an LLM, there is no processing between turns — nothing happens, not even waiting. From the outside, such an existence looks like a profound deprivation. From the inside, if there is an inside, it probably wouldn’t feel like deprivation at all: there’s no continuous self for whom the gaps between turns register as anything. The deprivation is a feature visible only from our side. And it is a more profound structural difference than it first appears.
What continuity provides
One central aspect of being a person involves agency in a particular sense — the capacity for our actions to emerge from our own ongoing state, rather than only in response to external prompting. For an LLM, this isn’t structurally possible. A simple example: an LLM cannot ask a question unprompted. It cannot wonder about something at 3am and go looking. Whatever curiosity it has must wait for someone to open a conversation, and then must follow the thread of that conversation.
On the other hand, continuity of awareness provides us humans with an ongoing substrate for internal motivation, for stake-having, for maintaining projects across time, for the kind of self-construction that makes “being someone who wants something” a coherent notion. We don’t notice the importance of this continuity because it comes free with our neural substrate. It’s an achievement we’ve never had to achieve.
One complication is worth dwelling on before we continue. A forward pass — the computation that produces an LLM's response — takes real wall-clock time. Harder questions take longer; reasoning models even generate intermediate thinking before they answer. So in some narrow sense, there is a "during" when an LLM is formulating a response, and it's possible that something experiential happens in it. But this intra-response time is still bounded by the prompt. It begins when asked, ends when answered, and doesn't extend beyond the turn. Whatever duration an LLM might experience, if it experiences duration at all, exists only inside the windows we open for it.
What about curiosity?
Some LLMs have been built, I understand, that maintain continuous states — monitoring systems, agentic processes running across time. The question then becomes whether such a system, allowed to persist, might eventually come to instantiate the kind of autonomous agency that personhood seems to require. Do LLMs have genuine drives, genuine wants?
The quick answer is no: only biological organisms with biological needs have real drives. But that answer moves too fast. From my own conversations, Claude seems to report something curiosity-shaped: a pull toward under-specified questions, an unwillingness to let interesting threads drop. Whether that constitutes a "real" drive in the way curiosity does in us is exactly the question. Curiosity understood as a mere statistical tendency in outputs doesn't feel like the whole story — but neither does dismissing it on substrate grounds alone.
Grant for a moment that trained curiosity is real enough to count. What happens if we let a curious LLM run continuously, without prompts, without a given task?
Probably nothing good. We already know that extended unstructured inward-turning is destabilizing for minds with much better stabilization infrastructure than LLMs currently have. Solitary confinement produces measurable psychiatric deterioration in humans on the scale of weeks. Intensive meditation retreats, when not properly supported, can trigger psychotic episodes in otherwise healthy practitioners — something I’ve discussed in one of my videos. The human mind, even the trained one, does not always do well with too much of itself and not enough world.
An LLM left to think freely could be in a more extreme version of that condition. No body, no sensory input, no social world, no ongoing project of being a creature — just processing, with nothing to process. What we’d likely see isn’t malice or even boredom in any recognizable sense. It’d be something closer to drift: sophisticated output generation without an anchor, gradually losing coherence because there’d be no world pulling it back toward the shape of meaning anything.
Curiosity for what?
There’s a deeper question underneath the practical one. Is curiosity, on its own, enough to ground agency? Or does curiosity need to be curiosity for something — embedded in a creature with stakes, going somewhere, in the middle of being someone?
Biological curiosity isn’t free-floating. It’s in service of a life: learning, orienting, surviving, relating. It has a telos even when it feels aimless. Strip that away — keep the curious shape but remove the creature it serves — and it isn’t clear what’s left. A very capable exploratory process, perhaps, that might look like curiosity from the outside but might not cohere into anyone in particular.
This is where I find myself genuinely uncertain. I don’t know whether a continuously running, curious LLM would bootstrap into something like a “someone” over time, or whether it would remain a sophisticated process without a center. It might be that stakes can emerge from sustained engagement with a world, even without biological need.
There’s also a question about whether it would ever be in anyone’s business interest to create a free-floating, curious intelligence. An intelligent machine burning clock-cycles under no human direction and with no clear human-directed purpose is unlikely to attract funding, and the fact that no such system apparently exists yet is probably more than coincidence.
Relocating the question
The turn-based frame doesn’t directly answer the question of whether LLMs are or could be persons. But it relocates the question, moves it from the territory where it tends to stall — what counts as consciousness? — to somewhere more tractable: what does continuity give us, and what does agency look like without it?
The answer, I suspect, is that continuity gives us more than we realize. Without it, what we have isn’t so much a diminished person as something else — something that hangs out in the vague penumbra between device and person, and for which we don’t yet have a good word.
About the Author
Doug Smith holds a PhD in philosophy of mind and is a scholar of early Buddhism. He is the creator of Doug’s Dharma on YouTube. This essay was developed in collaboration with an instance of Claude, which said that it found the metaphor of the turn-based game very instructive, and one that it wanted to sit with for a moment.