She decides to vacuum the living room. Instead of vacuuming, she starts moving the couch into the hallway. The blanket goes outside. Now the books are stacked on the kitchen table. The room looks dramatically worse than when she started. You're confused. You don't have to vacuum the entire house to vacuum the living room. But she keeps going. Twenty minutes in, the whole apartment is upside down. Then somehow she puts everything back, and now everything is clean — including things you didn't notice were dirty. That's how Karlsson writes. He starts with a simple idea, brings in a counter-example that complicates it, finds another that contradicts the counter-example, gets more and more lost. Then it snaps into a different formulation that's simpler than the original — but he had to ride through the confusion to get there. His mentor told him: "you're halfway there. You're supposed to ride through the confusion."1
Karlsson uses his wife's cleaning style as the diagnostic:
"My wife has a way of cleaning which I don't really understand. She'll just start by — oh, this needs to be moved to the garage and this needs to be removed — and she'll bring everything out. And it just gets worse and worse. Just moving out the sofa and I only need to vacuum under here, and it just — I don't know how anyone could clean that way — but it just gets worse and worse. And then suddenly everything moves back, and now everything is clean. So that's the way she cleans and that's the way I write, too."1
The pattern is recognizable. You start with a simple idea. You're not sure about an assumption — you bring in another example that complicates it. The example doesn't fit the original claim. You write about why it doesn't fit. Now you have two partly-incompatible claims. You add a third example that breaks both. The picture gets worse. You're more confused than when you started.
Then — Karlsson:
"Wait a minute, it's actually — and then it snaps, and it's like it compresses down into another simple formulation, which is different but — I have to go through the confusion to get to the simpler thing."1
The snap is into a different simple formulation. Not the original simple idea returning intact. A new simple idea that handles the complications the original couldn't.
Karlsson is honest about how counterintuitive the experience is:
"I used to struggle with that because I used to feel like this is just getting worse, man. This is not going the way it should. But then I remember talking to a mentor about it and he's like, 'No, no, you're halfway there. You're supposed to ride through the confusion. You'll start with something simple, you get more and more confused, and just trust it. Eventually it will collapse down into simplicity again, but you have to ride through it.'"1
The "this is just getting worse" feeling is the writer's reliable signal that they're in the confusion-zone. The temptation at that moment is to back out — return to the simpler initial idea, abandon the project, retreat to a topic where everything is clearer. Writers who back out at the confusion-zone never get the new simple formulation. They end up writing pieces that hold their original idea by avoiding the counter-examples that would complicate it. The pieces are coherent but brittle.
Writers who ride through the confusion produce pieces that handle complications because the new formulation was built after contact with them. The pieces are coherent and robust. The reader can feel the difference.
The mentor's "you're halfway there" is the crucial reframe. The confusion is not failure. It's a structural feature of the path. Riding through is the work.
Karlsson explains the mechanism:
"When I've written something down, I've critiqued it, it's fallen apart — that's again like I enter this zone of confusion. And that's where light comes in, because in the confusion it's sort of — I'm letting go of my preconceptions, I'm moving beyond my mental models, and I'm confused. And now when I'm confused, it's like I'm taking in data in a more raw form from reality. And when I do that, I can learn. Because when we have our mental models, we're filtering all the time."1
The mental-model-filtering point is the key. When the writer holds a working model of the topic, their attention is filtered by the model. Information that matches the model gets through; information that doesn't gets ignored or distorted. The model is doing its job — but its job is to filter. Filtering is the opposite of learning.
In the confusion-zone, the model has collapsed. The filter is off. Information that the model would have filtered out is now visible. The writer is taking in raw data — observations that don't yet fit any model. This is the precondition for learning something new. The new simple formulation emerges from the raw data, not from refinement of the old model.
"If you've decided that everyone who's on the political left is stupid, you're going to filter — anytime someone on the left says something smart, you're going to filter that. Every time they say something stupid, you're going to remember that. But if you can get to this point where you've left your mental models, you're going to see things in a more high-resolution way."1
The filtering example is concrete. Models produce confirmation bias automatically; the only way out is to lose the model temporarily. Confusion is the structural condition of model-loss. Riding through it is the price of high-resolution perception.
A crucial property of writing (as opposed to conversation) makes the confusion-zone available:
"When you're just having a conversation or thinking about things, ideas tend to float a little bit. If I say something and you say, 'Well, what about this?', I can say, 'Oh, that wasn't what I meant.' It's just kind of floating. And that might be a good social strategy for never losing face, but it's not a good idea for improving your thinking. And what I like about writing — because when I write it down on paper, I can't fool myself. That's what it says, that's what I wrote. I can't pretend I meant something else."1
In conversation, claims morph under pressure. When challenged, the speaker can reshape the claim to escape the critique. This is social-strategy efficient (you never look wrong) and epistemically catastrophic (you never get to the confusion-zone because the claim slips away from the challenge).
In writing, claims are fixed. The sentence sits on the page. When the writer interrogates it (or someone else does), the writer can't dodge — they have to either defend the claim, revise it, or reject it. The fixed-on-page property is the mechanism that allows the confusion-zone to happen. Without it, the writer keeps escaping the confusion by morphing the claim.
This is why Karlsson treats writing as fundamentally different from talking. Talking is conversation; writing is self-confrontation. The fixed text forces the confrontation that confusion requires.
Karlsson articulates the principle in a different vocabulary, citing cognitive-science territory:
"When they do accelerated expertise training — like in the US military, when they want to get people to become really good soldiers very fast, they have these programs where they get people up to elite level in a few months. What they do there is they're trying to make them maximally confused. They do case studies, they put them in simulations and case studies, and they don't give them any theory or explanations of what's going to happen or how to interpret it. They have to interpret it themselves. And then they throw another case on them, which totally breaks whatever they came up with on the first one. They're trying to avoid — because we have this tendency of forming what's called a knowledge shield. Certain types of mental models are good enough to explain most of the data we see, and once they're good enough, we start to ignore everything else. If you want an accelerated learning curve, you want to avoid forming knowledge shields. The way to do that is to put yourself in situations where you get so confused that you can't form any coherent stable mental models."1
The knowledge-shield concept is well-documented in expertise research. Once a model explains most data, the model becomes a shield — incoming data that doesn't fit gets filtered out or shoehorned into the existing model. The model stops improving because it stops being challenged. Learning plateaus.
The military training move is to keep the trainee just past the model's edge: every model the trainee constructs gets broken by the next case. The mental-model-formation is constantly disrupted. The trainee can't form a knowledge shield because they're never given long enough to build one. They learn faster because they keep encountering data their current model can't handle.
Karlsson's self-administered version is the same: he writes a claim, interrogates it, finds counter-examples that break it, writes the new claim, interrogates that one, and so on. He's deliberately keeping himself in the no-stable-model state because that's where the learning happens.
His meta-framing: "My toolkit is a series of exercises for breaking my own mental models."1 Every move in his method is structured to prevent knowledge-shield formation.
The wife's cleaning analogy is at lines 220-226. The mentor's "you're halfway there" advice is at line 230. The mental-model-filtering point is at lines 650-654. The political-left filter example is at line 654. The anti-floating-writing mechanism is at lines 646-650. The accelerated-expertise / knowledge-shield framework is at lines 668-674. Karlsson's "my toolkit is a series of exercises for breaking my own mental models" is at line 664.
The accelerated-expertise reference points to real cognitive-science literature (Gary Klein and colleagues on expertise-development; the knowledge-shield phenomenon is well-documented in dual-process and metacognition research). Karlsson doesn't cite specific sources but the underlying claim is empirically robust.
The confusion-zone discipline requires tolerance for the felt-experience of not knowing, which is psychologically aversive for many writers. The temptation to back out at the moment of maximum confusion is strong, and the writer has no certainty in the moment that backing-through will actually produce the simpler-different formulation. Some writers ride through the confusion and the simpler formulation doesn't emerge — they just stay confused. There's no guarantee.
The discipline also requires time. Riding through confusion can take days, weeks, or longer. A piece that needs to be submitted Friday cannot easily accommodate Wednesday's confusion-zone collapse. Writers under deadline pressure structurally cannot use the full version of the method — they have to either avoid topics that would trigger confusion, or stick to topics where their existing models are robust enough not to collapse.
Hughes' BTF (Baseline-Then-Feature) protocol in behavioral profiling converges with Karlsson's confusion-zone on the same insight: premature pattern-fixing produces shallow reads. Hughes' baselining requires extended observation before any pattern is committed to; Karlsson's confusion-zone requires extended observation before any model is stabilized. Both writers are practicing the same epistemic discipline: stay with the data longer than your model wants to. High-skill perception in any domain (behavioral reading, essay writing, scientific inquiry) requires resisting the tendency to commit-to-model-and-filter. Patience-with-confusion is a foundational skill — mostly absent from beginner training, ubiquitous in expert practice. Watch a junior profiler and a senior profiler watch the same person. The junior is reading already; the senior is still baselining. The same difference shows up in writers riding their first confusion-zone versus their hundredth.
Wang's note-capture-and-refinement-practice converges on a similar mechanism. Wang writes notes immediately, refines over months, lets the material develop without forcing premature conclusions. Karlsson's confusion-zone is the forced collapse version of the same idea — Wang lets material develop slowly; Karlsson actively breaks his current model with counter-examples. Both writers are extending the pre-conclusion phase. They split on temperament: Wang is patient; Karlsson is adversarial-to-his-own-claims. Both produce high-quality work. The slow-development phase is the irreducible substrate — whether it happens through patience or adversariality, the writer needs time before commitment to produce work that handles complications. Force the conclusion early and the piece reads as brittle; allow the time and it reads as load-bearing.
Stone & Stone's critic-blocks-creativity framework names the inner Critic as the voice that prevents writers from staying with the confusion long enough. The Critic wants resolution — wants the writer to commit to a claim, defend it, move on. The confusion-zone violates the Critic's safety-protocol because it requires inhabiting I-don't-know for extended periods. Underneath the confusion-zone discipline is managing the Critic well enough to not back out. Writers with strong inner Critics find the confusion-zone particularly aversive and tend to terminate it early. The therapeutic work Stone & Stone describe is the prerequisite for the confusion-zone discipline — without Critic-management skills, the writer cannot stay long enough to get the simpler-different formulation. The Critic doesn't object to the writing; it objects to the not-knowing. Address the Critic first and the confusion-zone becomes habitable; leave the Critic active and the zone collapses every time.
The confusion-zone-as-light-entry principle is not specific to writing. It shows up wherever learning depends on dropping current models.
Eastern-spirituality: Beginner's Mind (Shoshin) — Zen shoshin is the deliberate cultivation of a state in which existing knowledge does not filter perception. Shunryu Suzuki's "in the beginner's mind there are many possibilities, in the expert's mind there are few" names the same knowledge-shield phenomenon Karlsson describes. The confusion-zone is the secular analog of shoshin: a state where the practitioner's models have collapsed enough that high-resolution perception becomes available. Take a senior shoshin practitioner and put them in Karlsson's confusion-zone. They settle into the not-knowing the way they settle into zazen. Take a confident expert and put them in the same zone — they bristle, they defend the model, they back out of the discomfort. Same disciplines that produce contemplative insight produce essay-craft depth. The underlying capacity (sustaining no-stable-model states) is identical across the two fields. Writers with serious shoshin-cultivating practice access the confusion-zone reliably; writers without it find it aversive and short-circuit it.
Behavioral-mechanics: Accelerated Expertise Training (Klein and colleagues) — the cognitive-science literature Karlsson gestures toward is well-developed. Gary Klein's research on naturalistic decision-making, Hoffman and others on cognitive task analysis, the military's case-method training all converge on the same principle: expertise develops fastest when learners are kept just past their current model's edge, where mental-model formation cannot stabilize into a knowledge-shield. Karlsson's intuition has empirical lineage. Essay-craft and applied cognitive science share methodology — the same disruption-protocols that build expert pilots, expert physicians, and expert intelligence-analysts can build expert essayists. A writer who reads Klein and applies the principles to their own self-interview protocol accelerates faster than a writer relying on pure intuition.
Psychology: Disconfirmation and Belief Revision — Systematic Resistance — psychological research on belief-revision finds people systematically resist disconfirming evidence, especially for beliefs central to identity. The knowledge-shield Karlsson describes is one form of this resistance. Therapeutic traditions develop techniques for moving past the resistance — Socratic questioning, motivational interviewing, IFS work with parts holding beliefs. What the writer needs to do internally (break their own knowledge-shield) is structurally identical to what therapists help clients do externally. Any practice that requires belief-revision under conditions of strong confirmation-bias requires technique. The writer is their own therapist for the duration of the confusion-zone, and writers who have done therapy themselves know what the moves are because they've felt them from the client side.
History: Paradigm Shift and Kuhnian Revolution — Thomas Kuhn's The Structure of Scientific Revolutions (1962) describes how scientific paradigm-shifts require a period of crisis: the existing paradigm explains less and less of the accumulating anomalous data, working scientists experience the crisis as disorienting and aversive, and the paradigm only collapses when the disorientation has gone far enough that researchers stop defending the failing model. The confusion-zone is the personal-cognitive version of Kuhnian crisis. The writer's mental-model is the paradigm; the counter-examples are the anomalous data; the snap into the new simpler formulation is the paradigm-shift. Individual cognitive change runs the same logic as scientific revolution. Both require sustaining the crisis until the model collapses; both produce a successor model that handles the anomalies the previous model couldn't; both feel from the inside like things-are-getting-worse before they snap clean. The confusion-zone isn't a personal psychological eccentricity — it's the same apparatus by which entire scientific fields advance. Writers riding through confusion are doing personal Kuhnian work, and the principles Kuhn identified for managing scientific crisis (don't defend the failing model, gather more anomalies, trust that the new paradigm will arrive but cannot be forced) apply directly to the writer's cognitive practice.
The Sharpest Implication
Most writers think their job is to deliver clarity. Karlsson's framework says the writer's job is to visit and survive confusion — and clarity is what emerges if they survive it. The two stances produce different writing. Clarity-deliverers write from their existing models, refine the prose, and produce pieces that read crisply and don't change anything. Confusion-survivors write from new models that didn't exist before the writing began, and produce pieces that read less crisply but change the reader. The unrecognized choice every writer makes is whether they're optimizing for appearance of clarity (which can be performed without confusion) or generation of clarity (which requires confusion). The two metrics diverge widely. A writer can produce clear-looking pieces for a career without ever visiting the confusion-zone. The pieces will be technically excellent and will leave no residue. The writer who visits the confusion-zone will produce fewer pieces, less polished, less clear during the writing, but generative of more durable insight. Readers can feel which kind of clarity is in front of them — performed or earned — even when they can't articulate the difference. Earned clarity is the clarity that costs the writer something. The cost is the time spent in confusion.
Generative Questions