There is a kind of intellectual work that feels effortless. It moves through existing structures without friction, produces results that fit neatly inside the current conversation, and is rewarded immediately. Call it smoothness. It is not trivial — it requires genuine skill, pattern recognition honed over years, and the ability to operate fluently within a mature paradigm. It is what experienced professionals do well.
There is a second kind that looks, from the outside, almost indistinguishable from confusion. It generates friction. It does not fit the current frame. It asks questions that sound naive or unnecessarily large. But when it works, it does something smoothness cannot: it compresses. Fewer assumptions carry more weight. Arbitrary constants fall away. The explanation reaches further with less machinery. Paul Dirac called this beauty, and he meant something precise by it — not aesthetic pleasure but explanatory efficiency at the level of structure.
These two qualities are routinely conflated. Both produce confidence. Both look like mastery. But they are structurally different, and the difference matters enormously for how organisations think, hire, and decide.
The smoothness filter works by detecting the absence of friction — and treating that absence as evidence. This is the crucial mechanism, and it is easy to miss. The filter does not merely reward fluency; it uses fluency as a test. If a new idea sits comfortably within the existing landscape of assumptions, if it produces no resistance when slotted into the current frame, that very smoothness is experienced as confirmation. It must be right because it fits. Philosophers of knowledge have a name for this structure: coherentism — the view that beliefs are justified by how well they hang together with other beliefs, rather than by any independent contact with the world. The problem, which epistemologists identified long ago, is that a perfectly coherent system can be entirely disconnected from reality. The system validates itself. When an organisation uses smoothness as its primary filter, it is, in effect, doing institutional coherentism: internal fit substitutes for external truth, and the question of whether the frame itself is correct does not arise, because the frame is what defines correctness. This is not stupidity. It is a highly optimised heuristic for working within a settled paradigm. The cost is that it is structurally incapable of recognising genuine novelty, because genuine novelty is friction by definition. The frame, as always, cannot see itself.
Dirac’s filter operates differently. His famous distinction is worth understanding carefully: the good engineer, he argued, knows which small variables to ignore — and ignores them deliberately, artfully, in full awareness of what is being set aside. This is a positive and demanding skill, not a shortcut. It requires understanding the structure of a problem deeply enough to know which simplifications preserve the essential behaviour and which would corrupt it. The good engineer is not ignoring things because they are lazy or overconfident; they are making a precise, conscious, calibrated decision about what the situation can bear. That is a different kind of expertise from smoothness, and it looks nothing like it from the outside.
The honest tension here is that Dirac’s kind of thinking — foundational, compressive, structurally ambitious — produces both genuine insight and elaborate nonsense in roughly equal measure. From the inside, they feel identical. Dirac himself knew this, which is why he was ruthless about discarding his own ideas. The danger is not in pursuing foundational beauty but in using it as self-justification for work that is simply unfinished. Thomas Kuhn’s account of paradigm shifts captures the structural context: revolutionary science looks wasteful and disruptive at the moment of transition, and only retrospectively does its efficiency become visible. But Kuhn also documented how many proposed revolutions simply failed.
There is a more unsettling version of the local-versus-global problem than any optical illusion can capture. Consider what happens not when a single project is badly designed, but when an entire profession, industry, or generation of strategists shares the same set of assumptions. Each individual analysis is locally coherent. Each strategy document passes every internal test. Each model is technically sound within its own terms. The assumptions do not feel like assumptions — they feel like the background conditions of serious work, the things you take for granted before the real thinking begins. No one is being tricked. There is no deliberately constructed paradox. There is simply a frame so widely shared that it has become invisible, and beneath it a global structure that, examined from outside, does not add up in the way the local analyses suggest it does. This is harder to see than an illusion precisely because there is nothing strange-looking about it. It looks like rigour. It looks like consensus. The problem only becomes visible at the macro level, and by the time the macro level forces itself into view, the damage is done.
This is the pattern behind a class of business failures that continue to surprise people who should not be surprised. The organisations that fail most dramatically are rarely the ones that lacked skilled people or execution capability. They are, more often, the ones that had both in abundance — and applied them with great precision within a frame that was quietly becoming wrong. Kodak did not fail to understand photography. Nokia did not fail to understand mobile phones. The people running those organisations were accomplished, serious, and well-informed. What they lacked was not competence within their frame; it was any reliable mechanism for detecting that the frame itself was shifting. Their internal systems produced smooth, coherent, locally valid answers to the wrong questions. The failure was predictable in retrospect and invisible in prospect, which is exactly what the local-global problem predicts.
The reason it continues to be a surprise is structural, not psychological. Two of the main mechanisms organisations use to protect themselves — hiring and strategy — both replicate the problem rather than solve it. Hiring filters for smoothness by design: the selection process rewards those who operate most fluently within the current frame, which means it systematically filters out the people most likely to notice when the frame is wrong. Those people generate friction. They ask questions that don’t fit the agenda. They are experienced, at interview and in the early stages of a role, as unclear, underdeveloped, or simply difficult. The people who pass the filter are, by selection, the ones least equipped to see what the frame is missing. Over time, and especially as organisations grow more successful within their current paradigm, this selection effect compounds. The room fills with people who are excellent at the existing game and structurally blind to the possibility that the game is changing.
Strategy processes replicate the same error. Most strategy reviews are coherentist by design: they test proposed initiatives against the existing portfolio of assumptions and reward fit. A proposal that sits comfortably within the current frame — that can be evaluated against existing metrics, compared to existing benchmarks, and approved by people who already understand why it matters — passes. A proposal that requires rethinking something upstream fails, not because anyone decides to reject it, but because it generates friction at every stage of the process. The strategy review, like the hiring process, checks whether the steps are well-built. Neither has a procedure for asking where the staircase goes. In the absence of that procedure, the global question is never posed — and an organisation full of talented people, running well-designed processes, can march coherently toward a cliff while every local indicator reads green.
The question is not whether to value smoothness or foundational thinking. Both are necessary. The question is whether your organisation has any mechanism at all for distinguishing them — and whether, having made that distinction, it has preserved the capacity to act on the answer. The people who would notice that the frame is failing are precisely the ones the system is designed to exclude. By the time the problem is obvious to everyone, the ability to respond to it has usually already been optimised away.
Further reading:
↳ Why the frame cannot see itself — the epistemological trap that makes smoothness invisible as a limitation.
Garden notes
- Why the frame cannot see itself — the smoothness filter is an instance of the frame’s inability to detect its own limits.
- Overoptimisation — selecting for smoothness is a specific form of narrowing to the measurable and rewarded.
- What certainty proves — the confidence that smoothness generates reveals as much about the frame as about the subject.
- Observation and reasoning — Dirac’s distinction maps onto the two modes of inquiry and the productive tension between them.