“As we move from selling tokens where they provide the entire reasoning chain, to selling automated white-collar work—an automated software engineer, you send them the request, they give you the result back, and there’s a bunch of thinking on the back end that they don’t show you—the ability to distill out of American models into Chinese models will be harder.” – Dylan Patel (20260314)
There is a recurring claim across otherwise unrelated disciplines — linguistics, epistemology, economics, cognitive science — that converges on the same structural insight: knowledge is fundamentally implicit. Not merely that some knowledge happens to be tacit, but that implicitness is the default condition of knowledge, and what we call “explicit knowledge” is always a partial, after-the-fact extraction from a much larger implicit substrate.
Daniel Everett calls this substrate the “dark matter of culture.” It consists of the ineffable and the unspoken — hunches, posture preferences, emotional dispositions, ways of seeing that are constructed through emicization rather than instruction. Unlike overt knowledge, which can be found inside and outside our brains (Wikipedia, textbooks, databases), dark matter is found only within the individual. It cannot be uploaded, copied, or transferred wholesale. Each person’s dark matter is the sediment of their particular trajectory through particular cultures, and it is this — not the explicit facts they can recite — that constitutes the bulk of what they know.
This is why culture is an implicit theory for its members. It dictates not just what to look at but how to reason about what one sees. The Charcot epigraph captures it: “In the last analysis, we see only what we are ready to see, what we have been taught to see.” Culture does not hand you a set of propositions; it hands you a set of dispositions. Explanation — explicit or inexplicit — precedes both your options and choices. Before you can even formulate a problem, your implicit theory has already constrained which problems are visible to you.
Language operates the same way. Language assumes shared knowledge, just as conversation is implicit in the context. Every sentence presupposes an enormous body of unspoken agreement about what things mean, what matters, and what can go without saying. Speaking the same language doesn’t mean two speakers share the same exact meanings — because meaning is not in the words. The explicit is built upon the inexplicit. This is why it almost never happens that two minds hold precisely the same idea, and why you can’t really transfer your knowledge to others — each piece of knowledge has to be created individually.
David Deutsch formalizes this from the epistemological side. Explicating the inexplicit amounts to knowledge, and knowledge is unpredictable, and never derived mechanically. Knowledge creation is not extraction — it is not pulling something pre-formed out of a warehouse. It is conjecture: knowledge must first be conjectured and then tested. The inexplicit is the raw material, but the act of making it explicit is a creative, fallible, unpredictable process. This is precisely why knowledge is by definition unpredictable — if you could derive it mechanically, it would already be explicit, and therefore not new knowledge at all.
The Austrian economists arrive at the same structure from an entirely different direction. Mises writes that man reveals only a part of his value scale through actions. “We deduce the existence of a specific value scale on the basis of the real act; we have no knowledge of that part of a value scale that is not revealed in real action.” Value — which is to say, subjective knowledge of what matters — is implicit. It can only be partially surfaced through behavior. Value scales are ascertainable from actions, not vice versa, because value scales cannot be exhausted. The parallel to Everett’s dark matter is exact: just as culture’s implicit knowledge cannot be fully articulated, economic preferences cannot be fully enumerated. What you do reveals a sliver of what you know you want.
This is the deeper point behind the competence-performance distinction. We can only figure out what people know by what they do, because we can never directly study competence — only performance. The Turing test is one such behavioristic mistake: exams can’t exhaustively measure a student’s competence. But performance is one of the few indirect windows we have. The error is not in using performance as evidence; the error is in confusing it with the thing itself.
If knowledge is fundamentally implicit, several consequences follow. First, the flexibility of humans is found particularly in the tacit knowledge — not in what we can articulate, but in the vast body of know-how, disposition, and dark matter that we embody without being able to state. Second, if you don’t create an explanation, you will be enmeshed in the situation — the implicit will think for you, through cultural defaults and inertia, unless you do the creative work of making some of it explicit. Third, this is why the best is seeing the front by yourself — because the inexplicit and unconscious ideas that matter most exist between heads, not within them, and no report or summary can substitute for direct encounter.
The insight is not that we should try to make everything explicit. To assume that we will eventually do away with the implicit disrespects the substrate-independence of information — the implicit could always be otherwise. Rather, the insight is that knowledge creation is the ongoing, never-completable process of explicating slivers of an inexhaustible implicit substrate. Every explanation, every price, every utterance, every performance is a partial surfacing. The substrate remains.
Related:
- 10-2g You can’t extract the meaning from the words only, because the explicit (seen) is built upon the inexplicit (unseen)
- 10-2h Explicating the inexplicit amounts to knowledge, and knowledge is unpredictable, and never derived mechanically
- 10-2g1 Culture is an implicit theory for its members. It dictates what to look at and how they reason about the world.
- 1-1a2a5 Overt knowledge can be found inside and outside our brains (e.g., Wikipedia), but dark matter is found only within the individual
- 13-6 Man reveals only a part of his value scale through actions
- 13-8a2c Value scales are ascertainable from the real actions, not vice versa—because value scales cannot be exhausted
- 10-2g1f3 We can only figure out what they know by what they do (‘performance’), because we can never directly study what people know (‘competence’). To assert that we can is a common error in thinking.
- 9-1b You can’t really transfer your knowledge to others, because each knowledge has to be created individually
- 5-2b4 Explanation (explicit or inexplicit) precedes both your options and choices
- 1-1c2a There is no escape we are and will be shaped by anti-rational memes
- Daniel Everett
- David Deutsch
- Ludwig von Mises
- Murray Rothbard