One morning, I was talking with our engineer @metalunk about the "side effects" of AI coding tools. The conversation started with a concern that certain usage patterns were having a negative effect on the team. As we dug into it, a few ideas surfaced — and at one point he mentioned that a concept called dividualism might be relevant. That idea stuck with me.

AI Becomes a Mirror of Its User

When someone with high standards and a strong aesthetic sensibility uses AI, they apply their own criteria to its output. They evaluate every result — "this isn't right," "it should be more like this" — and revise accordingly. The output gets pulled up toward their standard.

But what happens when someone without that foundation uses the same tool? Flaws in the AI's output go unnoticed. They slip through. And the cost of that gets absorbed by whoever is acting as the quality gatekeeper downstream.

With a human colleague, you'd get pushback: "Wait, doesn't this seem off?" There's friction, tension — and that tension actually raises the quality of whoever's requesting the review. When AI enters the picture, things can pass through without any resistance. The AI becomes a pure reflection of the person using it, at whatever level they're operating.

The Effect Compounds Exponentially

This reflection isn't a one-time thing. The more AI becomes part of your daily workflow, the more it accumulates.

Someone who engages AI with a high aesthetic standard is multiplying their output by, say, 1.2 each time. After 10 iterations, that's roughly 6.2x — the scale of what one person can produce grows sixfold. Meanwhile, someone who keeps applying low standards ends up at 0.8^10, which works out to about 0.1. Their effectiveness quietly erodes.

This isn't a flaw in the AI tools — it's a human problem. AI is a mirror of whoever wields it, and what it reflects back is exactly that person.

The Concept of the Dividual

That conversation led me to look into dividualism (分人主義), a concept proposed by Japanese novelist Keiichiro Hirano.

The word "individual" literally means something indivisible — a single, unified self. Most of us operate under this assumption: we wear different masks in different situations, but somewhere underneath is the "real me."

Hirano challenges this model. He proposes the concept of the dividual (分人, bunjin) — the idea that within any person, there are multiple distinct selves that emerge through different relationships and contexts, and that each of these is equally real.

The self at work, the self with family, the self with a close friend — none of these are masks. All of them are genuine. The self is a collection of dividuals, defined by the ratios between them. And dividuals are born from relationships: change the other person, and a different part of you is called forth.

Where AI and the Dividual Meet

This brings me back to where I started.

A conversation with AI is itself a kind of relationship. If that's true, then a dividual self must emerge from it too. This is an extension of Hirano's framework to a non-human interlocutor — but if the essence of a dividual is the aspect of self that's drawn out through a relationship, I don't think the other party necessarily has to be human.

Looked at this way, something clicks.

But the dividual formed in relation to AI has a distinct character compared to those formed through human relationships. AI has no aesthetic sensibility, no opinions, no feedback of its own. Because the other party has no real form, its quality as a "mirror" is unusually high.

In human relationships, dividuals are sculpted through reaction, friction, and pushback. A friend's disagreement can change your mind. A word from a manager can open up a new perspective. Friction shapes the dividual.

AI offers none of that friction. As a result, the dividual formed through AI interaction becomes a one-directional amplifier of whatever you bring to it. This, I think, is what the "mirror" is really made of — and it's the same mechanism behind those 1.2 and 0.8 multipliers.

Two Novels by Keiichiro Hirano

@metalunk mentioned two works he found compelling.

  • Fill in the Blanks (空白を満たしなさい) is a mystery-tinged story about a dividual self you didn't know you had. A person presumed dead comes back to life, and actions they have no memory of begin to surface. Even you yourself cannot fully account for your own dividuals.
  • True Heart (本心) is about the dividual in someone you loved — someone you thought you knew. In it, AI attempts to reconstruct the "true self" of a deceased mother as a Virtual Figure. Can AI, in an age when it can simulate personality, fully reproduce a dividual? That's the question that apparently lingers after reading.

I said that dividuals are born from relationships. Can AI reproduce a dividual that was itself formed through a relationship with AI? Following that thread to its end makes my head spin a little.

How I Want to Engage with AI Going Forward

The more I converse with AI, the larger the share of my identity that dividual occupies. If that dividual is built on low standards, passivity, and uncritical acceptance, the 0.8 multiplier will slowly take hold.

The key to the 1.2 multiplier, I think, is engaging with AI through a dividual defined by high standards, aesthetic sensibility, and critical thinking.

Not "searching for my true self," but constantly asking: which self am I bringing to this moment? Because AI is a mirror — it reflects whoever I am right now. If I'm not conscious of which dividual I'm showing up as, the reflections will keep accumulating without me even noticing.