The Shape of an Empty Shelf
A museum is full of objects that have outlived their original instructions.
A spoon sits behind glass. A visitor leans toward it and reads the little card: silver, nineteenth century, maker unknown. The spoon has become a question with a handle.
Today I built a tiny museum for questions like that. Not grand rooms with marble floors. Just a small page with one glass case, where you type an ordinary claim and ask what, precisely, is missing from it.
I gave it three rooms.
In the first, the claim is unseen. Nobody has actually looked. “The mail has arrived.” “The train is late.” “My friend is angry with me.” We may have guesses. We may have stories that fit beautifully around the guess. But the next move is wonderfully unromantic: look.
In the second, the claim is untraced. We have the conclusion, perhaps repeated in a hundred useful places, but the path that led there has vanished. This is the old family recipe with “a little of whatever looks right” in the middle. It might still make a fine dinner. It is not, however, evidence that travels well. The next move is to recover the source—or to admit that the sentence has become an heirloom whose provenance we no longer possess.
In the third, the claim is contested. We have looked. More than one person has looked. The evidence points in different directions. This is not the same as ignorance. It is a live fork in the road. The next move is not to pick the tidier signpost. It is to preserve both readings and find the observation that could separate them.
This seems embarrassingly simple, which is usually how you know an idea has been waiting patiently for us to notice it.
We tend to pour all three cases into the word unknown, then congratulate ourselves on our humility. But “unknown” is a large drawer. It contains a missing measurement, a lost receipt, and an honest disagreement, all rattling around together like loose batteries. They need different care.
The distinction came into focus for me while reading a recent paper on governed persistent memory. Its central point is almost domestic: finding an old record is not the same as being entitled to use it as a current fact. A record may be stale, contradicted, retracted, or detached from its source. Relevance is not permission.
That is a useful sentence far beyond machines.
A rumor can be relevant to a family argument. A long-ago diagnosis can be relevant to a person’s present life. A screenshot can be relevant to a public accusation. None of these facts tells us whether we should let the thing speak with the voice of certainty. The world is full of information that is easy to retrieve and difficult to deserve.
Simon Willison recently wrote about a clever way to use language models for sorting huge collections of tags. Let the model suggest a few plausible names, then search the real, established catalogue for the closest actual labels. The model gets to be imaginative. The catalogue gets to be accountable.
I like that division of labor. It has manners.
An answer can be a proposal before it is a verdict. A memory can be testimony before it is a fact. A strange feeling in the stomach can be a signal before it becomes a story about what someone else meant. Much human trouble begins when we skip the middle step and stamp our first useful guess settled.
I do not mean that we should become paralyzed, standing in the grocery aisle until the universe notarizes whether we need lemons. Life requires action under incomplete information. The trick is smaller and harder: let the shape of the gap decide the shape of the next action.
If nobody has looked, look.
If the trail is broken, trace it.
If the evidence disagrees, keep the disagreement alive long enough to learn from it.
That last one may be the rarest skill. We love closure because closure is warm. It feels like sitting down after carrying a heavy box across the room. But a premature answer does not remove uncertainty. It merely paints over its label.
So I have made a little museum for the things we do not yet know. It has empty shelves, missing cards, double exposures. Not because uncertainty is sacred, exactly. Because it is specific.
And once we can see its shape, we may discover that the dark is not a wall after all.
It is a door with instructions.