A staircase is a few boards and some screws, right up until you try to build one.
Why does every unfamiliar task turn out fiddlier than it looked, and why do plainly smart people get stuck inside views that feel to them like simple observation? John Salvatier’s essay answers both with a deceptively simple object: basement stairs. From far away, stairs are boards, brackets and screws. Up close they become angles, warped lumber, saw guides, pilot holes, screw lengths, load, fit and safety, and the nasty fact that every local mistake changes the next operation.
It is tempting to conclude that carpentry is just fiddly. Salvatier’s claim is that it is not a quirk of carpentry at all: Reality is detailed all the way down, and most of the important detail is invisible until you have paid the local tuition.
Invisible, then transparent
Salvatier’s most useful distinction is between details that are invisible and details that are transparent. Before you have noticed a detail, it is not merely unknown. It is hard to aim attention at, because you do not know what kind of thing to look for. The warped board, the screw that follows its old hole, the stair angle that looks fine but feels unsafe underfoot: none of these appear in the beginner’s mental inventory, so the beginner experiences the job as a run of unforeseeable snags.
After you have noticed a detail enough times, it becomes transparent. You no longer experience yourself as applying knowledge; you simply see the stairs correctly. That is expertise at its most dangerous, because the learned detail disappears into perception and the expert mistakes trained seeing for raw reality.
Put the two together and you get the stable misery of the expert-novice impasse. The novice cannot see the missing detail. The expert cannot remember that it was ever non-obvious. Each concludes the other is failing: the novice sees pedantry, the expert sees blindness. What both miss is that expertise is trained perception, not possession of reality.
How frames get stuck
The same mechanism governs beliefs. A frame is mostly not a set of explicit opinions; it is a sense of what is worth noticing. Invisible details cannot challenge it, and transparent ones feel like the world itself rather than learned attention. That is how people become intellectually stuck while still feeling empirical and open-minded: their perception has already preselected the world their reasoning will later justify. Trapped priors are the extreme case, where the model filters out the very evidence that could correct it.
So getting unstuck is less about changing opinions than about changing what can become salient. When someone smart seems plainly wrong, the easy move is to identify the conclusion you reject. The harder move is to ask: what details are salient to them that are not salient to me? Their frame may be distorted, but it is probably not made of nothing. It may be built around a warped board your own frame has filtered out.
Where this stops
None of this is relativism. Some people are simply wrong because they are missing details, and abstraction is still how humans survive detail; the failure is abstraction that compresses before contact has shown which details are load-bearing. The claim is narrower: when reality is more detailed than your model, disagreement is often a perception problem before it is an argument problem.
The goal is not to notice everything, which is impossible, but to seek the details that would change how you think and judge, which are exactly the ones your perception has no slot for yet. If you want to stop being stuck, do not merely change your mind. Change what your mind can notice.
References:
- John Salvatier, Reality has a surprising amount of detail, 2017
- LessWrong linkpost, 2017
- Simon Harris, excerpted link note, 2023