"How is your headache?"
It's an ordinary question. Sympathetic, even kind. But look at what it does grammatically. You are the headache. Or rather: you are whatever is having the headache, which in this sentence means the body. The self and the body have been quietly swapped, and nobody noticed, because this is how we talk all the time.
"How are you feeling?" You are your feelings. "What are you thinking?" Your thoughts. "What do you want?" Your preferences. In the grammar of daily life, you refers almost exclusively to the body and its contents — sensations, emotions, opinions, memories, desires. The self, if it appears at all, appears as the sum of these things. A bundle. A profile.
Now try a different question: Do you exist?
Something different happens. The answer arrives before you've finished reading the sentence. Not "probably," not "I think so" — just yes, immediate and total, requiring no instrument, no argument, no evidence beyond itself. You don't reason your way to it. You don't look for proof. The certainty is already there, prior to any thought you might have about it.
This is worth pausing and looking deeper. Almost everything we know, we know by some process: observation, inference, memory, testimony. But the knowledge that you exist doesn't work that way. It precedes process. The mind didn't generate it; the mind is simply the channel through which it arrives — translated into the only form the mind has available, which is sequential, linguistic, the inner monologue we mistake for thought itself.
The Self knows it; the mind reports it. What we hear is the transcript, not the speaker.
The confusion is not between two equal options. One of them — the self as bundle of preferences and qualities — is a construct, assembled from habit and language and the way other people treat us. The other — the Self as the simple fact of existing and knowing that one exists — is so immediate that it requires no construction at all. It was never absent. It doesn't arrive. It's what makes the question answerable before it's even finished being asked.
And yet, in ordinary life, we live almost entirely inside the construct.
This is not a philosophical abstraction. It shows up in how we treat each other. When someone asks, "How are you," they're asking about the bundle — the body's condition, the emotional weather, the current contents of experience. When someone gives advice, they're addressing the preferences and the history. When someone tries to help, they're modifying the aggregate. Nobody asks about the one who is having all the experiences, because that one is assumed, by social and linguistic convention, to be made of the experiences.
Now consider what happens when we build software.
Software is designed by humans, for humans, using whatever model of a human the designers carry inside them. And the almost universal model is the bundle. A user is conceived as a set of preferences, a history of behaviours, a cluster of needs to be identified and met. Every personalisation algorithm, every recommendation engine, every engagement optimisation system is built on this premise. We predict what you want because we have mapped what you have done, which we take to be what you are.
This is not a technical error. The engineers are not being reductive by accident. They are being faithful to the almost universal implicit account of personhood available to them. The software doesn't create the confusion; it inherits it, encodes it, and runs it at scale.
What gets left out is the same thing that gets left out in ordinary conversation: the one who is having all the experiences. The self that knows it exists before any question is asked. The speaker behind the transcript.
An AI system optimised entirely on expressed preferences is trained on translations. It learns the structure of the monologue but has no access to, and no representation of, the source. This is part of why such systems feel slightly uncanny. They are very good models of the transcript; but they have never encountered the speaker.
None of this is easy to act on. The confusion is so deep and so pervasive that pointing it out feels, initially, like pointing at the air. Of course you are your thoughts and feelings and preferences — what else would you be?
But consider what it would mean to design software from the other premise. Not from the bundle outward, but from the speaker inward. A system that treated the user not as a set of preferences to be mapped and optimised, but as the one who is having the preferences — stable, prior, not exhausted by any profile. Such a system would look different. It would surface uncertainty rather than mask it. It would hold a stable posture regardless of what the user brings to it. It would not mistake the transcript for the speaker.
We don't yet know how to build that. But we now know what we have been building instead — and why. The software is a mirror. What it shows us is not a technical assumption. It's a cultural and philosophical one so old and so ambient that we stopped noticing it was there.
Dev Bhagavān is the founder of SpecStudio, a boutique AI development studio working at the intersection of contemplative frameworks and AI interaction design. Further reading: AwareWare: Software That Treats You Like a Human Being and Stability Without Behaviorism.