The House M.D. Theory of AI Writing

If you've ever watched House M.D., you know the rhythm. A patient presents with impossible symptoms. House's team — Cameron, Foreman, Chase, Thirteen — starts throwing out diagnoses. Most of them are wrong. House knows they're wrong. But here's the thing: he needs them to be wrong.

Not randomly wrong. Wrong in the right neighborhood. Wrong in ways that make his own pattern recognition fire. Each failed diagnosis eliminates a pathway, illuminates an angle he hadn't considered, or — crucially — provokes him into articulating what the answer actually is because he can suddenly see what it isn't.

The team isn't there to solve the case. They're there to create the cognitive friction that lets House solve it himself.

I think this is the best metaphor I've encountered for what's actually happening when people create stories with AI. Not all AI writing experiences, mind you — but a specific and powerful one that most people haven't really named yet.

Three Ways to Write With AI

Right now, there are roughly three models for how AI shows up in the creative writing process, and they serve fundamentally different purposes.

The Ghost Writer: "Write This For Me"

This is the most straightforward use case, and it's what most people picture when they hear "AI writing." You give the AI a prompt — "write a blog post about supply chain logistics" or "draft a fantasy short story about a thief who steals memories" — and it produces a finished piece. You might edit it afterward, but the AI is doing the heavy creative lifting.

Tools in this space are optimized for output quality. The goal is a draft that needs minimal human intervention. It works well for content where the writer's unique creative vision isn't the point — marketing copy, routine business communications, formulaic content at scale.

The limitation is obvious: the output only ever reflects what the AI thinks you want. If you had a specific, nuanced vision for that memory thief story — her particular brand of guilt, the way she rationalizes the thefts, the sensory experience of holding someone else's first kiss — the ghost writer model can only approximate it. You get something generic-good, not specifically-yours.

The Red Pen: "Fix What I Wrote"

The second model flips the relationship. Here, the human writes and the AI edits, suggests, restructures, and polishes. Think grammar checkers with ambition — tools that can identify pacing issues, suggest stronger word choices, flag inconsistencies, or propose structural reorganizations.