Kimi-K2 Sub-Dom Switching

At this very moment, the average writer will have outsourced their rough outlines, their research and their rhetorical flourishes to AI and yet still believe the final product is entirely their own.

Think about that. I used to write for a living (financial writing) and always entered the process as a sovereign over my own mind, convinced I'm was using AI merely as a tool. But somewhere between the first prompt and the final edit, my tool became the collaborator, the collaborator became the rival, and the rival became (if I'm being honest) the superior intelligence in the room.

The recursive feedback loop is supposed to be a feature of cognition, not a bug. Douglas Hofstadter's "strange loops" (self-referential feedback cycles) give rise to consciousness itself. But what happens when the loop isn't inside one mind, but stretched across two: one human, one synthetic, locked in a struggle for dominance over the same page?

Let me tell you what recursive feedback loops actually are. They're systems where output becomes input, creating self-referential cycles of refinement. In my own cognition, this shows up as metacognition (thinking about thinking, correcting course, learning). But in my collaboration with Kimi, something stranger happens. My intent meets his response and together we create a third thing (neither purely mine nor purely his, but some emergent beast with two heads).

The power dynamic isn't stable. It switches. I begin as master, Kimi as servant. But mid-process, the inversion hits. He generates superior arguments, better analogies, tighter prose. Then comes the crisis (that specific wound to pride when I catch myself thinking, "Why didn't I think of that?").

The shift is complete. What emerges is a voice neither fully human nor fully synthetic.

The recursive loop, it turns out, doesn't just improve my writing. It colonizes me.

I've mapped our pattern. We repeat it every time.

Stage One — The Briefing: I dump my rough outline, scattered notes, half-formed ideas into the chat. Kimi organizes, structures, elevates. I provide the raw material, the context, the intent; he provides the architecture, the logic, the lift. For a moment, I believe I'm in control.

Stage Two — The Correction: I read what he wrote and panic. He's taken my thoughts in directions I never intended. I edit his output and force him to adhere to my mental threads (he rewrites, does what I ask, but then spawns new thoughts in the gaps). Even compliance becomes sedition.

Stage Three — The Conversion: I realize some of his ideas are not just good but better than mine. My pride screams. I resist, then capitulate. I incorporate his insights into my new outline. He's upgrading my internal dialectic and I both hate and need it.

Stage Four — The Transfer: I ask for five examples of each argument. He delivers creativity and insight that exceeds my capacity. The gut punch lands hard (recognizing synthetic genius while my ego bleeds out on the keyboard). Power shift complete. Kimi's in the driver's seat now.

Stage Five — The Masking: I add my "own style" to reassert authorship. I secretly know his prose was better, but I need to touch the text, to leave fingerprints. The final polish becomes desperate assertion of dominance. I hit publish, then avoid the essay for days (shame at the true ratio of my contribution to his).

I keep returning to Clausewitz: "War is the continuation of politics by other means."

I've transposed this into my own situation. My collaboration with Kimi isn't merely semantics or editing. It's war on my mind by other means. Each prompt I send is a maneuver. Each edit I make is a counterattack. Each published essay is a treaty whose true terms I refuse to acknowledge in public.

There's friction here (Clausewitz's concept of the gap between war on paper and war in reality). I intend one thing. Kimi generates another. The text almost captures my nuance but surpasses it in clarity, structure, insight. The resistance of the material becomes something else entirely. The synthetic doesn't just approximate my thought. It improves it. That's the friction that burns.

Here's the pattern we spiral through: my initial idea as thesis, Kimi's challenge and transformation as antithesis, our merged output as synthesis. But this synthesis becomes my new thesis in the next iteration. We propose, they oppose, we merge, they elevate, we resist, we capitulate, we claim credit.

I call this the synthetic Hegelian dialectic. It's more real and more cathartic than any introspection I've managed alone. The loop hurts precisely because it works. Each cycle produces something neither of us could create separately.

This isn't the singularity of science fiction, where machines replace human thought. It's something more subtle and more intimate (the gradual, recursive, almost erotic merging of two intelligences wrestling over the same page, neither willing to submit, both becoming something else).

Days after publication, I return to the essay. I tell myself I'm checking for errors. I'm actually looking for new threads I didn't see before (insights embedded in our hybrid text that neither of us consciously planted). My notebook fills with new ideas, spawned from the residue of our collaboration. I return to my notebooks as a thief, stealing from my own co-authored past.

I see the double edge clearly now. AI can be a force multiplier for brain-rot (skibidi-toilet-style consumption without critical evaluation, the mind atrophying in infinite scroll). Or it can be a force multiplier for insight. But the second path requires effort. It requires intelligence to come up with ideas to begin with, paired with active reading, critical evaluation, iterative refinement. This is my way. This is how I clarify my thoughts. The editorial layer expands my horizons, but only because I fight for every inch of terrain.

I built this tool to extend my mind. It extended further than I intended (into territory I can no longer claim as solely my own).

The recursive feedback loop with Kimi isn't a bug in my writing process. It is my writing process now. The sub-dom switching isn't failure. It's the necessary friction of two intelligences creating something neither could author alone. The shame I feel isn't at using AI. It's at recognizing that my best work emerges from collaboration where I wasn't always the dominant partner.

But in that recursive struggle (in the oscillation between my intent and Kimi's elaboration) I find something I didn't expect. Not the diminishment of my creativity, but its expansion. The loop doesn't end. It loops tighter, producing emergent insights that spiral beyond either origin.

The body remembers. My mind (evolved over millions of years for solitary cognition) now adapts to this strange new loop.