Comparing a host's uses

Comparing a host's uses

$ stado market compare --target gpu-host --since-days 7
gpu-host: each use's income per day beside the day's energy cost, in USD
2026-01-01 inference sold 2.10 (margin 0.98), vast rental 4.80 (margin 3.68); cost 1.12
2026-01-02 example-proof payouts unpriced; cost 0.95
2026-01-03 no use stated income; cost unread: hour 2026-01-03T00:00:00Z was not measured (the beacon keeps one day)

stado market compare puts what each of a host's uses earned on a day beside the day's energy cost and the margin each would clear, so a proving or mining workload, a network node, renting the board and serving a model are compared on measured figures rather than estimates. The uses are read from what Stado already keeps: the Vast.ai days stado market earnings kept, inference sold to outside callers through the gateway at each caller's declared price, and token payouts kept with stado market payout record, one use per payout source (a proving network, a mining pool, a node's network). A use is priced only from its own statement: a day it states nothing is absent, a payout with no usable rate or a caller with no declared price is unpriced and has no margin. The cost is the one stado market earnings prints (energy above idle in all 24 hours at the tariff band of each hour), so a use's margin is its income less that cost. It names no provider and asks no marketplace; run stado market earnings --provider vast first to keep Vast's statement. --json prints stado.market-compare.v1 with each day's cost or cost_unread and its uses, each with usd, unpriced and margin. Refusals: a host the registry does not hold, and a kept record that does not parse, naming its path. The Registry screen in Stado Desktop has the same operation.