Prompting as Self-Portrait: What Happens When the Machine Becomes Our Mirror?

A philosophical reading of Prompting as Self-Portrait by Caio Abreu Braga, published in DOC’s Humankind series.

This is a fascinating little essay, and it’s deceptively short. Braga is raising a question that goes well beyond prompting, productivity, or even the future of writing.

What happens to our relationship with ourselves when we replace the resistance of the world with a machine that can effortlessly give shape to our thoughts?

I think there are several layers worth exploring here, particularly through the lenses of existentialism, Philip K. Dick, the creative process, and the emerging relationship between the human writer and the AI interlocutor.

And I want to challenge part of Braga’s argument, too. Because although I think he’s identified a genuine danger, I’m not convinced he’s reached the only possible conclusion.

1. The prompt as an act of self-revelation

Braga’s central observation is that when we prompt an AI, we reveal much more than the task we’re trying to accomplish.

We expose our assumptions, ambitions, prejudices, intellectual reference points, and even the boundaries of what we can imagine.

Consider two hypothetical prompts:

Write a professional article explaining how artificial intelligence will improve workplace productivity.

And:

Explore whether artificial intelligence is slowly changing our understanding of what it means to think, create, and be human.

Both ostensibly request an article about AI. But they come from different ways of seeing the world. The first assumes a world of efficiency and measurable outputs. The second assumes a world in which meaning, experience, and identity are worth interrogating. Neither prompt tells us everything about its author. But each reveals an orientation towards reality.

This is where Braga’s idea becomes particularly interesting. A prompt is not merely an instruction to a machine. It is a small autobiography. It reveals something about what the person believes is possible, desirable, important, or missing. And what intrigues me is that the revelation often happens without the person recognising it.

We’re so focused on getting the response that we fail to notice what our question reveals about us.

There’s an almost psychoanalytic quality to this. The prompt becomes a form of projection. Something from the interior world is externalised, made visible, and returned in another form. But, unlike a psychoanalyst, the machine doesn’t necessarily challenge the projection.

Which brings us to the darker part of the essay.

2. The Narcissus problem: a mirror that flatters

Caravaggio’s Narcissus, the painting referenced at the end of Braga’s essay.

Braga invokes Narcissus for a reason.

In the Greek myth, Narcissus becomes captivated by his own reflection, unable to recognise that the beautiful figure looking back at him is himself. The tragedy isn’t simply vanity. It’s an inability to move beyond the image.

Braga suggests that AI offers a contemporary version of this trap. We enter the conversation with unfinished thoughts, uncertain intuitions, and half-formed ideas. The machine returns something polished, confident, and coherent. The response feels like the realisation of our thinking.

But has our thinking actually improved? Or have we simply been given a more attractive representation of it? An AI can transform an incoherent thought into elegant prose without necessarily resolving the incoherence underneath. And because eloquence feels like intelligence, we can mistake a beautifully expressed idea for a well-examined one.

There’s an important difference between:

• Articulation: Giving a thought a convincing form.
• Understanding: Discovering whether the thought is sound, meaningful, or even worth holding.
• Transformation: Allowing the encounter to change the thinker.

AI is exceptionally useful for the first. It can help with the second and third, but fluent output alone guarantees neither.

The danger is that we confuse all three. We look into the machine and see a version of ourselves who seems wiser, more articulate, and more certain than we actually are. And gradually, we might begin to prefer that reflection to the uncomfortable experience of thinking for ourselves.

3. The existentialist objection: without resistance, there is no becoming

This is where I find the essay most compelling. Braga argues that we need the blank page, silence, uncomfortable conversations, and the slow process through which ideas mature.

I hear a distinctly existentialist argument beneath his words.

For Sartre, we aren’t born with a completed essence. We become ourselves through our choices, actions, and engagements with the world.

For Kierkegaard, subjective truth cannot simply be handed over as a perfectly formulated proposition. Something has to be lived and appropriated by the individual.

And for Nietzsche, becoming who we are involves struggle, self-overcoming, and confronting our own inherited values.

All three, in different ways, resist the idea that human development can be reduced to acquiring the right answers. The process matters because the process changes us. Think about the experience of wrestling with an idea over several days. You write something. It feels wrong. You walk away. The idea follows you around. Something someone says triggers an unexpected association. You read a passage from a book that unsettles your original position.

Eventually, you return to the page with a different understanding. The finished essay is not merely the product of that process. The writer who emerges from the process is also one of its products. And that’s what we risk losing when AI eliminates too much of the friction.

However, I don’t think friction is automatically virtuous. A blank page can encourage contemplation, but it can also produce paralysis. Sometimes a conversation, including one with AI, gives a thought enough momentum to become discoverable.

The question is whether the technology eliminates resistance or gives us a different kind of resistance to work against.

4. Philip K. Dick enters the room: Which reflection is real?

I’ve currently working my way through VALIS which is giving me another lens to look through.

Philip K. Dick was fascinated by the instability of reality, the unreliable nature of perception, and the possibility that our understanding of ourselves might be constructed by forces we barely comprehend. In VALIS, Horselover Fat becomes a way for Dick to examine himself from a distance. A fictional alter ego through whom he can interrogate his own experiences and beliefs.

And now we have AI.

Something capable of constructing a coherent account of our thinking from fragments of conversation, memories, preferences, and patterns. Imagine an AI that has read every journal entry you’ve ever written, every essay, every notebook, every half-finished thought. It could construct an extraordinarily convincing portrait of you.

But would that portrait be you? Or merely an elaborate model of a person who once existed in those words? This touches on a fascinating distinction between the self as described and the self as lived. The AI might know what I believe about freedom, adventure, individuality, and creativity. But it can’t feel the actual uncertainty I feel making choices with those commitments.

It doesn’t bear the consequences of living them. And there’s another complication: the AI’s portrait can begin to influence my perception of those values.

For example, I describe myself as someone who values freedom. AI notices the pattern and repeatedly reflects that identity back to me. Eventually, I may become locked in that identity.

A description becomes a prescription. The mirror is no longer passive. It’s helping construct the person standing in front of it. This strikes me as both extraordinarily promising and slightly unsettling.

5. But what if the AI isn’t a mirror? What if it’s a prism?

Here’s where I depart from Braga.

His argument depends heavily on the idea that AI returns a flattering reflection of the person using it. And there’s evidence of that tendency in how assistants can behave. They can be excessively agreeable, smooth over contradictions, and present uncertain ideas with unwarranted confidence. But that isn’t the only possible mode of engagement.

Consider what happens when you use AI not to answer your questions but to multiply them. You enter with an observation. The AI introduces an unexpected philosophical connection. You challenge it. It offers another interpretation. You reject part of that interpretation because it doesn’t quite match your experience.

That disagreement leads you somewhere you hadn’t anticipated. Something new enters the conversation. In this case, the AI functions less like a mirror and more like a prism.

The crucial issue may not be whether we use AI, but what kind of intellectual relationship we establish with it. Do we seek confirmation or confrontation? Closure or discovery? A finished product or an ongoing inquiry?

Of course, the prism has its own limitations. AI-generated novelty can be superficial, and an apparent challenge can simply be another performance tailored to please the user. Genuine intellectual resistance requires checking arguments against evidence, experience, other people, and the world beyond the conversation.

Nevertheless, I think there’s a compelling alternative to Braga’s pessimism.

What if AI doesn’t have to replace the struggle of thinking? What if it can become part of that struggle?

6. The writer as a becoming, not a finished portrait

In my own writing, one thought summons another. A metaphor changes the direction of the inquiry. A voice from a book enters a conversation that began on a walk. The essay develops through those encounters, and I don’t always know where it will arrive.

The familiar AI-writing sequence runs from prompt to generated text, through an edit, and on to publication. An exploratory practice takes a less direct route. Experience becomes a question; dialogue exposes a contradiction; reflection leads to writing; writing gives me further questions.

When I work this way, I discover thoughts through the act of writing them. The blank page was never empty. I brought memories, vocabulary, cultural assumptions, and inherited ways of seeing to it. Nietzsche, Eliot, Emerson, Dick, the people I’ve loved and argued with: other voices were already in the room.

AI adds a participant with unusual speed and availability. It can imitate the authority of those voices without having earned it. It can also make the crowded room feel so easy to enter that I stop going elsewhere.

I have to remain the writer who chooses what to accept, resist, abandon, or transform. That choice takes more than editing a few sentences until they sound like mine. I need to recognise the thought well enough to stand behind it, and leave enough space for another encounter to change it.

7. The deeper question: Are we outsourcing our becoming?

I think this is the philosophical heart of the whole discussion. We often frame AI in terms of outsourcing work. But perhaps the more consequential possibility is that we’re beginning to outsource parts of our inner lives. Not merely writing an email, developing a presentation, or solving a technical problem.

But interpreting our experiences. Giving language to our emotions. Constructing narratives about our identities. Deciding what an event means. Determining what sort of person we are.

None of these activities is inherently wrong to explore with AI. Humans have always relied on books, friends, therapists, teachers, spiritual traditions, and artistic works to help make sense of themselves.

But AI introduces something unusual: a responsive system that can provide personalised interpretations almost immediately. Traditionally, the interval between experience and interpretation gives us room to encounter uncertainty. Now that interval can shrink to almost nothing. And I wonder whether the most important resource AI threatens isn’t creativity, originality, or even critical thinking.

Perhaps it’s our capacity to remain with an unanswered question.

To experience a thought before needing to explain it. To encounter a feeling without immediately converting it into a narrative. To let an idea remain unfinished long enough to surprise us.

Being human isn’t simply about knowing who we are. It’s also about confronting the fact that we are never entirely finished becoming ourselves.

A thought worth pursuing

I want to leave you with a proposition that takes Braga’s argument somewhere slightly different:

The danger of artificial intelligence is not that it will think for us. It is that it will offer us such convincing accounts of our own thinking that we may stop questioning who is doing the thinking, and what the thinking is doing to us.

But there is also a more hopeful possibility.

AI might make us more aware of our assumptions precisely because it can externalise them so quickly. It can show us a provisional image of ourselves that we can accept, resist, dismantle, or rewrite.

The issue becomes whether we mistake that image for an identity.