Why some inscriptions cannot be moved safely

About one inscription in thirty-five sits somewhere that makes spending it hazardous — not because anything is wrong with it, but because of where it happens to sit in its output. Measured across 140 randomly sampled cats.

Two situations make an inscription risky to spend. Neither is a defect, and neither is visible in a picture of it.

1. It is not at the start of its output

An inscription has an offset: how many satoshis into its output the inscribed satoshi sits. Usually that is 0. When it is not, spending the output means the inscription is somewhere in the middle of the value being counted out, and first-in-first-out decides where it lands. Getting it back to a known position takes a transaction built specifically to do that.

Real examples from this collection: one cat sits 48,696 satoshis into its output. Another is inside an output of roughly 4,666 BTC — an exchange's consolidated UTXO, where the cat is one inscribed satoshi among hundreds of billions.

2. The output carries other inscriptions

Several inscriptions can share one output. Spending it moves all of them, and their destinations are decided by the same arithmetic. One C.A.T. shares an output with 359 other inscriptions. Nothing that touches that output can move one without moving the rest.

How common, actually

This was worth measuring rather than guessing, and the first attempt got it wrong in an instructive way. Sampling the first 28 cats suggested roughly one in five was affected — alarming, and false. The front of the collection is disproportionately held by exchange addresses, and exchanges consolidate, so the sample was biased towards exactly the case being counted.

A proper random sample of 140 cats gave 2.9%: about one in thirty-five. The lesson generalises beyond this collection — take the sample randomly, because inscription numbers are not randomly distributed among holders.

What to do about it

If an inscription is at a non-zero offset, the fix is to isolate it: a transaction that puts it alone at offset 0 of its own output. Any competent inscription-aware wallet can do this, and it is the holder's decision rather than something a tool should do on their behalf while they thought they were doing something else.

What a tool should do is refuse, and say why. "This cannot be done safely" is a better answer than a transaction that technically succeeds and puts an irreplaceable artwork somewhere nobody intended.