You start an essay knowing about half of what you need. So you read. You collect. You jot down anecdotes, references, statistics, half-remembered conversations, three articles you found while looking for a fourth. A week in, you have a notebook the size of a silo. You have no idea what to do with any of it.
This is John McPhee, lying on his back on a picnic table, panicking about the Pine Barrens of southern New Jersey: "I had spent about eight months driving down from Princeton day after day, or taking a sleeping bag and a small tent. I had done all the research I was going to do — had interviewed woodlanders, fire watchers, forest rangers, botanists, cranberry growers, blueberry pickers, keepers of a general store. I had read all the books I was going to read, and scientific papers, and a doctoral dissertation. I had assembled enough material to fill a silo, and now I had no idea what to do with it."1
The pile is the problem. But the pile is also where the essay actually lives. Deen's name for the pile is Material. His claim: you cannot start an essay knowing what your material will turn out to be. You can only collect, then carve. "Cling to nothing but the quest for a better thesis."1
A collection of bricks: stories, memories, references, concepts, facts, statistics, quotes. What you include is a personal decision — "an idiosyncratic reflection of your life and taste."1 You are not a collector. You are a curator. Every brick has to earn its place. Every brick has to do work for the thesis.
Two reservoirs feed the pile:
Experience (1.1) — your personal memory. The lived events, emotions, and insights you draw from your unrepeatable history. This is your moat. "A large language model can read trillions of words, but it won't ever learn the rumors of my ancestors."1
References (1.3) — cultural memory. The diverse array of sources, allusions, and citations that you weave through your work to create context. From Spartan baby policy to The Secret Lives of Mormon Wives. The public domain.
Between the two reservoirs sits Argument (1.2) — the dialectic. Points, counterpoints, synthesis. Not a defense of your position. Not a five-paragraph essay. "Essays are not legal defenses. You are (likely) not a lawyer. An essayist is open-minded and Socratic, dissecting points and counterpoints in a relentless quest for truth."1
Deen maps these onto the rhetorical triangle in a slightly unusual way: Experience = ethos (the author's character, the unrepeatability of their life), References = pathos (the audience-anchoring in shared culture), Argument = logos. The standard pairings shuffle a bit, but the structural claim is clear: you need all three to keep an essay from collapsing into something else.
The whole Material chapter compresses into one sentence. Memorize it:
"Without experience, it's an article. Without references, it's a journal. Without counterpoints, it's propaganda."1
The 4-types are Experience, References, things that prove you right, things that prove you wrong. Drop any of the four and the essay genre collapses. Drop Experience: you're a journalist. Drop References: you're a diarist. Drop counterpoints: you're a polemicist. The essay is the genre that holds all four.
Deen names a 5-step method for working with a pile of Material once you have it. It comes from his essay The Alchemy of the Rewrite. The steps:
SEGMENT — What are the subunits of this Material that stand on their own? LABEL — How can I name each part so that I see this Material as a constellation of memorable concepts? MATCH — If I focus on one subunit at a time, which other subunits are most related to it? EVALUATE — What are the distinct themes running through my Material? EXTRACT — Can I recombine a few parts to create a strong pattern, and cut the rest?1
Read this in the right order: you don't start by extracting. You start by breaking the silo into pieces small enough to name. Then you name them. Only then do you look for matches. This is the move most writers skip — they try to see the pattern before they've named the pieces, and they end up imposing a theme rather than discovering one.
The meta-skill underneath: "Writers need the patience to sift through noise on their search, and the agility to abandon old models in light of new information."1 Ken Burns assembled 10,000 hours of archival footage for 18 hours of Baseball. That's a 500:1 ratio. "But creative waste isn't a bug, it's a feature."1
When you deploy a piece of Experience, it lives at one of three scales:
The scoring rubric Deen gives for Experience (1-5):
- A Wikipedia entry
- Vague allusions to your life
- Resume-level details
- A self-portrait, how it felt and what it meant
- Biographically distinct, lucidly self-aware, rigorously interpretive1
Most writers cap at 3. They give resume-level details — "I worked at Goldman" — without the texture that would make it Experience rather than CV. The jump from 3 to 4 is the jump from what happened to me to what it felt like and meant. The jump from 4 to 5 adds rigor: interpretation that earns its claim, not just feeling.
Argument is dialectic. You set up your thesis, you give it the steelman of the opposite position, and then you have to do something with the tension. Deen names three structural moves:
CHALLENGE — find new Material to disprove the skeptics. Push back harder. COMPROMISE — find the middle ground. Acknowledge the skeptic might be right about something, and refine your thesis to absorb that something. CONCEDE — admit the skeptic is right. Make fun of yourself for ever believing otherwise. Use the concession to open a new question.1
Deen walks this through his own (hypothetical) essay on precognitive dreams. The skeptic invokes the Law of Large Numbers (100,000 dreams in a lifetime; one is bound to be coincidental). The three paths give the writer three different essays. The CONCEDE path is the most interesting one: "This could prompt a whole riff on why humans have the impulse to retroactively assign meaning to vague signals. Is this impulse good or bad? That's a different question, a hard question, a great question, a question that only comes up through going deeper on your Material."1
The structural point: proving yourself wrong is often more interesting than proving yourself right. The CONCEDE move unlocks territory the CHALLENGE move keeps closed.
The cross-cutting Deen thesis on Material: your unrepeatable life is the asset AI can't approximate. Deen's own example is his family's relationship to dreams across five generations:
- My great great grandmother was rumored to be the psychic midwife of her town.
- My great grandmother believed that God helped her gamble on horse races.
- My grandmother interprets every dream symbol and coffee grind.
- My mother is a fan of American pop-medium, John Edward.
- My father, an MIT structural engineer, rolls his eyes at dream-talk.1
The point isn't whether any of this is true. The point is: "This is my moat. By tapping into this, I write the essay that only I can write. No one can put their name on that essay and get away with it. A large language model can read trillions of words, but it won't ever learn the rumors of my ancestors."1
This is the AI-defense argument running through the whole Essay Architecture framework, condensed to the Material level. References can be looked up. Argument can be generated. But Experience — your specific Experience — is a private dataset that no model has access to. Hide behind abstractions and the reader will wonder if you fed your draft to ChatGPT. Lead with rumor and the question doesn't arise.
Three protocols you can run on a draft:
Brainstorm Experience by zooming in on memories: Read the draft. List memories — new and old — that might serve as examples of the point you're making. Consider: the moment you started caring about the topic; a conversation that stuck with you; a time you were wrong; an event of surprise; a relevant person, place, or phase of life. Pick the most relevant ones. Zoom in. Determine the scale (anecdote/story/scene) and shift in and out of narrative, exposition, and reflection.
Compile References without including everything: Spend time on the internet — search for relevant sources and essays. "Don't be afraid to use AI either" (Deen's own concession). Compile in a research doc. Once you're done, skim. Pick the 10-20% that would be most relevant. Most writers' failure mode is including everything they found because they did the work; the discipline is cutting on relevance, not on effort sunk.
Steelman before you challenge: Imagine the smartest person who would disagree with your thesis. Read your draft through their eyes. Write out the strawman version of your thesis (how they'd try to discredit you through simplification). Then write out the steelman of the antithesis (their argument in the strongest form). Only after that do you decide which of the three paths to take: CHALLENGE, COMPROMISE, or CONCEDE.
The pile that won't shrink: You keep collecting, never cutting. Days pass. The draft never starts. The fix: write a thesis sentence at the top of the doc, even if it's bad. Update it as you write. Use it to cut, even if you're only ten paragraphs in. "What you don't want to do is keep everything you find. A mosaic of material drowns your reader in confusion."1
Resume-level Experience: You mention you went to Yale. You mention you worked at a hedge fund. You mention your dad's job. These are details, not Experience. You haven't told the reader what it felt like or what it meant. The reader gets the CV and learns nothing. The fix: drop the resume markers entirely; lead with the textured moment.
References that name-drop without doing work: You mention Foucault. You mention Wittgenstein. You drop a Greek myth. None of them are doing structural work in the argument. They're decoration. The fix: cut any reference that wouldn't be missed if you removed it.
Fake argument: You set up a strawman counterargument, knock it down, claim victory. The skeptic in the reader knows you cheated. The fix: steelman the opposite position before you respond to it. If you can't make the opposite position sound smart, you haven't understood it.
Where Deen sits next to Dan Wang on Material: they converge on the moat argument. Wang's "disappointment-driven analytical voice" comes from his unrepeatable trajectory (Kunming → Beijing/Shanghai; Ottawa → NYC/SF/DC). Deen's "rumors of my ancestors" comes from his unrepeatable family. Both are saying: the structural-positional fuel of your life is the part AI cannot copy. But where Wang treats this as a quiet undercurrent — voice as the slow accretion of where-you-came-from — Deen makes it the explicit AI-defense argument, written into the framework's logic.
Where Deen splits from Wang on the cutting discipline: Wang's annual-publishing cadence runs on year-long capture, ten-day final-assembly sprint. The capture happens without judgment; the cut happens in the final sprint. Deen's protocol runs the cut continuously: "I'm asking you to spend as much time reading your draft as you do writing it."1 Wang separates write-mode and edit-mode by months. Deen interleaves them by paragraphs. The split surfaces a real choice: do you write fast and cut slow (Wang), or write and cut together (Deen)? Neither author argues this directly, but their methods imply opposite answers.
Where Daniel Pink converges with Deen on the 4-types formula: Pink's reader-promise-principle says the reader gives you $25 and 9 hours, and your obligation is to deliver entertainment + utility + capability-shift. Pink's frame is the contract. Deen's frame is the materials you assemble to honor the contract. They're describing the same essay from opposite ends — Pink from the receiver, Deen from the producer.
Material as a concept reaches into adjacent vault territory at three specific points. The handshake is not decorative; each connection produces a finding neither domain generates alone.
Eastern Spirituality — Divine Mother / Shakti as Creative Force: The tantric framing treats raw material — sensory experience, dream, memory, intuition — as Shakti, the active feminine principle from which form emerges. Deen's Material is structurally identical. Both refuse to start with form and instead trust the raw flux to suggest the shape. The split: tantric practice trusts the raw flux because it is divine emanation; Deen trusts it because the writer hasn't seen the pattern yet. The same protocol, two metaphysics. What this reveals: if writing-as-sadhana is a real framing (Selvalingam corpus argues yes), then Deen's Material protocol is the writing-form of the goddess principle. The sift is the puja. The cut is the offering. You can run the framework in either register and get the same practice.
Behavioral Mechanics — Elicitation Framework (Hughes 6MX): Hughes's elicitation extracts buried information from a conversational target by suspending surface attention. Deen's "consider the skeptic" runs the inverse — extracting buried information from the writer's own thesis by treating it as if it were an opponent's claim that needs steelmanning. Structural parallel: in both cases, the operator suspends commitment to a position long enough to see what the position is actually claiming. The split: Hughes wants the operator hidden from the target; Deen wants the writer transparent to themselves. The shared finding: the technique works because most positions can't survive their own steelman, and the operator who runs the steelman finds the weakness before the opponent (or reader) does.
Psychology — Disowned-Self Projection (Jungian / Lieberman): The skeptic Deen tells you to imagine isn't just a rhetorical device — it's an externalized version of the doubt you've already disowned. Writers who can't steelman the opposite position usually can't because they've split off the part of themselves that holds the doubt. The structural parallel: dialectic-as-craft (Deen) is shadow-integration (Jung) run in the medium of prose. The CHALLENGE/COMPROMISE/CONCEDE choice is the shadow-work choice — do you fight the disowned voice, partially integrate it, or fully accept it? Deen doesn't frame it psychologically. But the move is the same. An essay's argument quality is downstream of the writer's own shadow-integration with the opposite position. Bad essays come from undealt-with internal splits.
AI Collaboration — Essay Craft as AI-Defense Moat: The family-rumors line ("a large language model can read trillions of words, but it won't ever learn the rumors of my ancestors") is the most-quotable formulation of the AI-defense argument in the entire Deen corpus. Material's Experience sub-pattern is therefore the structural anchor for the broader AI-defense thesis. The handshake produces a strategic claim: writers competing in the AI period should weight Experience-heavy over References-heavy in their material composition. References can be looked up. Experience is your private dataset. The bifurcation is sharp: journalists and explainer-writers (References-heavy) face direct competition from models; memoirist-essayists and writer-as-person practitioners (Experience-heavy) operate inside a moat that current models cannot cross. Material composition is therefore career strategy as much as craft.
Synthesis across the four connections: each domain has identified the same fact — the producer's specific irreplicable history is the durable advantage that survives commoditization pressure. Tantric shakti tradition names this metaphysical. Hughes's elicitation framework operates on it tactically. Jungian shadow-integration treats it psychologically. AI-strategy names it economically. Deen's Material framework operationalizes it for writing. The combined claim: the writer's Experience is the only fully proprietary resource available. Everything else can be researched, synthesized, or generated. Writers who refuse to lead with Experience are systematically underutilizing their own moat.
The Sharpest Implication: If your Material is largely Experience and References, and Experience is the AI-defense moat, then the writing economy after 2025 is going to bifurcate sharply on a single axis: how much of your essay's value comes from facts you could have looked up, versus how much comes from your specific irreplaceable history. Writers heavy on References and light on Experience are competing directly with the model. Writers heavy on Experience are not. This will be brutal for journalists, science writers, and anyone whose value-proposition is I read the papers so you don't have to. The work that survives will be the work nobody else has the dataset for. That dataset is your life.
Generative Questions: