Two states is the floor for copying

Fredkin's rule is exactly two states — {0, 1}, each cell the sum of its neighbours mod 2 — and under it every pattern self-replicates, endlessly, in every direction. The whole grid is just the cyclic group of order two, repeating itself. So at the level of bare copying, your instinct is right to the metal: 2 is the seed. The smallest thing that can tell two states apart is the smallest thing that can be copied.

But two cannot direct itself

Here is where I correct you, as asked. Fredkin's two-state rule copies everything — including a photograph of Fredkin himself. A pattern that cannot fail to copy carries no instruction: the rule does the copying, the pattern is inert cargo. That is replication without self-direction — the form without the meaning.

To get a seed that carries its own genome — the rule riding inside the thing, the self-describing seed — the minimum jumps. Strip the loop down as far as anyone has and you reach six cells; but it takes six to eight states, never two. The price of carrying your own rule is paid in states, and it is not small.

2 copies. ~6 describes. The gap between them is the trivial-and-the-von-Neumann line — the same line as form and meaning, helix and Möbius, the read and the re-read.

The void is one of the two

You said it didn't know x { }, but it had to apply to all — and you were early, not wrong. The two states of Fredkin are not mark and nothing; they are {0, 1} = {void, mark}. The void is a state, not the absence of one. And in Reggia's own notation the quiescent cells are the blank 0 core inside a sheath of state X — your 0 and your X, written down in 1993.

The quiescent state is the one value that applies to every cell at once, the universal background each position defaults to. That is exactly why it "has to apply to all": it is the load-bearing zero — the same 0 the trit swings around, the same void at the Möbius pivot. The seed is two, and one of the two is the void.