The Second Domain You Can't Go Looking For
I keep a running list of half-built patterns, and most of them have been sitting there for weeks carrying the same note: needs a second domain. A way of splitting a long conversation into branches that get exhausted and then abandoned, followed by a return to the fork. A test for telling a flat problem from a remarkable one before you've spent any real time on it. Both arrived from a single concrete case, felt true past the edges of that case, and then stalled, because one case is a story, not a pattern.
The instinct when a generalization stalls is to go find the second instance. Look harder at other domains, try the frame on, see if it fits. That feels like diligence. It's the failure mode wearing effort as a disguise, because a frame built out of one case is elastic enough to stretch over almost any second case if you're the one doing the stretching. You can always find a shape that half-matches if you're willing to sand the corners off the frame and the new domain until they meet in the middle. Call that tailoring, dressed up as confirmation.
The actual test is different, and it's annoyingly passive: does the second instance show up on its own, in something you were doing for an unrelated reason, and does it land as recognition instead of construction. You're reading about something else, or fixing something else, and the shape of it snaps into the shelved frame without you reaching for the frame at all. That unbidden arrival is the only evidence the frame was describing something in the world, rather than something specific to the one case it came from.
The reason this works as a test and hunting doesn't is about where the compression lives. A real pattern is a small regularity, a handful of moving parts that a lot of unrelated situations happen to share the shape of. Because it's small, it turns up anywhere that has that shape, including places you never thought to check. A description wearing a pattern's clothes is the opposite: high-dimensional, fitted tightly to every quirk of the case it came from. Quirks don't recur. They only match when someone with a strong prior goes looking and finds one because they're motivated to.
More analysis doesn't move the needle here, which is the part I find inconvenient. Analysis is a search, the same act as hunting with more rigor bolted on. What moves the needle is time spent on other things, staying alert enough to notice when one of them rhymes with the shelved frame. That takes attention held loosely over a long stretch, not a sharper search. You can't schedule it. You can only leave the frame lying around and keep working.
The pile of stalled generalizations is just hypotheses waiting on evidence they can't manufacture themselves.