Why Wang Captures Everything and Koe Captures Nothing
A Man Who Sells Notebooks Won't Use One
Dan Koe is building software whose entire pitch is capture your ideas. Sitting with his co-founders discussing that exact product, he mentions that he doesn't bother writing most of his own down — anything more than seven days old, he's not going back to, and if it actually mattered it'll surface again on its own.1 He used to worry about losing ideas. Now his position is: whatever, it'll come back.2
Dan Wang does the opposite, absolutely. All year, every year, sentences go into his phone the moment they arrive — listening to Mozart, eating a dumpling, noticing something in Beijing. Not ideas in the abstract. The exact sentence, because the precise phrasing is the thing worth keeping.3 The discipline is total: if a sentence is worth remembering it gets captured immediately, specifically to prevent the writer's standard tragedy of thinking I'll remember that and then losing it forever.4
Two working writers. Two systems that cannot both be followed. And the interesting part is that neither is being sloppy — each has a real mechanism behind it.
What Koe Is Actually Claiming
Strip the contradiction and Koe's argument is about where the filter lives. Instead of treating memory as leaky and the notes app as the safe, he inverts it: let your mind be the filter, and let it decide what's worth keeping by what it bothers to resurface.5 Important things recur. They come back days later, often improved. If a thought never returns, it probably wasn't load-bearing.
The cost he's protecting against is specific. Reaching for the phone is a small act of evaluation and task-switching — stop, judge the thought worth saving, switch devices, type, switch back. Each break costs presence. You traded one preserved idea for the three downstream ideas the unbroken stream would have produced.6 And there's a second cost: constant capture trains the mind to stop holding anything, so it quits doing the background work of turning a thought over and pushing the good version back up.
That's not a claim about laziness. It's a claim that forgetting is doing real work — a filter that runs for free and clears the trivial automatically.
What Wang Is Actually Claiming
Wang's system isn't the opposite of Koe's on the question of trust. It's the opposite on the question of what the notes are for.
Wang's phone notes are not a memory aid. They're raw material for a single annual object. The whole year is a gathering phase pointed at one essay, and the notes are the input pile the December sprint distills.7 He isn't capturing so he can remember; he's capturing so that in eleven months there's enough accumulated sentence-level material for a structure to emerge from. The app is explicitly not precious — it's a bucket, the writing in it unpolished.8
Which means Koe's central objection doesn't quite land on him. Koe's worry is that capture breaks a creative state. Wang's capture isn't happening during a creative state. He's got a day job doing financial writing; the letter's real cognitive work happens in the refinement layer over months, and later in the ten-day sprint.9 The phone note is a two-second act during ordinary life, not an interruption of flow.
The Variable That Actually Sorts Them
Here's what decides it, and it isn't temperament or discipline. It's whether you currently have a frame worth filtering for.
Koe's mechanism runs on resurfacing — ideas come back if they matter. But what makes an idea "matter" enough to resurface is your active project and current frame; the vault names the underlying mechanism directly, and it's goal-tuned.10 Change your project and different old ideas start floating up. Which means trusting your mind to keep what matters is only as good as the frame doing the keeping. A writer with no meaningful project has a filter resurfacing junk, and trusting that mind just returns junk with confidence.
Wang has the strongest possible version of a frame: one essay, one subject, one hard date. His capture is pre-filtered by a commitment that's been running all year, which is why the pile coheres in December instead of scattering.
Koe's instruction, read precisely, is a fast, not a policy — the no-capture rule sits inside a seven-day protocol for recovering a fried, overstimulated creative state.11 It's a treatment for a specific condition: too much input, no digestion, the mind trained out of doing its own processing. Wang doesn't have that condition. He has a frame and a deadline and eleven months of quiet.
So the honest sorting is: capture when you have a project the capture is feeding; fast from capture when you're trying to recover a creative state you've lost. Those are different situations, and a writer in one situation reading advice written for the other will follow it straight into the failure mode.
The Part That Stays Unresolved
Koe contradicts himself and the vault preserves it rather than smoothing it over. His creativity advice says trust forgetting; his business advice says build the swipe file, hoard the inputs, never lose a proven format.12 The charitable read — delivery mode versus discovery mode — is exactly the distinction this report just drew, but Koe never states it himself, and the vault is explicit that whether it's a coherent two-mode system or an unexamined inconsistency is left open on purpose.13
The genuinely unanswerable question underneath all of this: how many good ideas does the forgetting filter quietly kill? You can't remember what you forgot, so the false-negative rate is structurally invisible.14 Koe's "important ideas resurface" is plausible introspection, not evidence — and it's the kind of claim that will always feel confirmed, because the only data you ever see is the ideas that came back.
Confidence
Koe's position, including the founder-versus-his-own-product framing and the seven-day protocol context, is [VERIFIED] against the vault's page, tagged [PARAPHRASED] [POPULAR SOURCE] at origin with timestamped line references.1 Wang's system is [VERIFIED] against a separate transcript-sourced page, different practitioner, different lineage.3 The Koe-versus-Koe contradiction is documented in the vault, not constructed here.12 The sorting variable this report proposes — that having a project-frame is what decides which system applies — is synthesis, drawn from the vault's own note that forgetting is only safe when the frame doing the filtering is worth it.10
Threads: → Why Wang Captures Everything and Koe Captures Nothing — Thread Map