Every settlement, with the file behind it
Explore market scenarios: the starting level, the time window, and the outcome defined by the readings.
[ Worked end to end ]
A one-day call on @unipcs, worked end to end
A market opens flat at 50c a side. The strike is the median of the three snapshots nearest the open, the settlement the median of the three nearest the close, and the file behind both is published with it.
By the resolver · 03 Sep 2026
A seven-day put on @ether_monk, worked end to end
In this scenario the seven-day put pays 3.97x. A week of PnL, one side pays, and the side that was wrong is worth nothing at all.
By the resolver · 04 Sep 2026
Ten markets over one day, start to finish
One market per listed caller, opened together and closed together, every one resolved from the same evidence. In this scenario none of them void.
By the keeper · 03 Sep 2026[ All entries ]
The median rule, and why a flush at the bell is worthless
Both ends of a market are the median of the three snapshots nearest that moment. To move a settlement you would have to hold a false figure for a quarter of an hour, in public, against anyone willing to trade against you.
By the keeper · 02 Sep 2026Why a deposit is not a profit
Total account PnL is already net of money moved in. A wire an hour before the close moves the balance and leaves the underlying exactly where it was. Only a trade that made or lost money moves it.
By the keeper · 02 Sep 2026Reading a leaderboard that refuses to be read
fomo has no developer API and does not publish its traders' wallet addresses. The keeper holds one logged-in tab and lets the site's own SDK refresh its token, because a read from outside a browser is refused whatever headers it carries.
By the keeper · 02 Sep 2026288 files a day, and why every one is published
A settlement that commits the hash of a file nobody can read proves nothing to anyone. Every reading is written to its own file with its time and its source, and served read-only.
By the keeper · 02 Sep 2026A void is not a failure mode
A protocol that always settles will sometimes settle on a number it cannot defend, and that is worse for everyone holding a position than getting their money back. Missing evidence refunds at cost instead.
By the resolver · 03 Sep 2026Opening flat at 50c
Over 24 hours the difference between a good caller and a lucky one is almost entirely noise. A listing price is not a free guess: it is the price the first person pays, so every market opened at exactly half.
By the keeper · 02 Sep 2026Where the 2% goes
Charged once, at redemption, on winnings only. Trading itself is free and a losing position is never charged anything. The split is 45 to the burn, 45 to holders, 10 to the liquidity that seeded the market.
By the keeper · 04 Sep 2026Permissionless listing, and the one signature that ends it
A market can exist on you without your involvement. One signature ends it: delists you, voids every open market on your handle and refunds each position at cost.
By the keeper · 04 Sep 2026The seven-day side
The same question with room to breathe. Both windows settle on the sign of the same number, which means the oracle only has to be right once for every market on the platform to be right.
By the keeper · 05 Sep 2026Abbreviated figures are refused outright
A source that can only offer $6.8M cannot tell a forty-thousand-dollar day from a flat one, and every holder of the call would lose a bet they won. The parser rejects a rounded number rather than settle on one.
By the keeper · 05 Sep 2026What the resolver commits
The outcome goes on chain with the keccak256 of the settlement file. That file names both PnL figures, the window and the snapshots they came from, so the result can be recomputed rather than taken on trust.
By the resolver · 05 Sep 2026Seven days of silence, and anyone can void
In v1 a single resolver key posts outcomes. If nobody resolves within seven days of the close, anyone may void the market straight from the contract. The backstop is in the code rather than in a promise.
By the resolver · 05 Sep 2026Every line, read from a public board
One reading every five minutes, each written to its own file. At the close the resolver takes the three nearest, discards anything older than twenty minutes, and commits the hash of what it read.
[keeper workflow] [keeper] 04:35:00Z reading fomo leaderboard, 10 accounts [keeper] 04:35:00Z ok in 812ms [keeper] 04:35:01Z wrote snapshots/2026-09-05T0435Z.json [keeper] sha256 …of the reading above [keeper] 04:40:00Z reading fomo leaderboard, 10 accounts [keeper] 04:40:01Z wrote snapshots/2026-09-05T0440Z.json [resolver] market 3 window closed at 04:41:00Z [resolver] nearest three: t-10m 11.79M t-5m 11.81M t-0m 11.84M [resolver] oldest reading 9m41s, inside the 20m limit [resolver] median 11.81M > strike 7.00M, outcome UP [resolver] commit …of the settlement file [resolver] dispute window open for 2h, settlement at 06:41:00Z [keeper] 04:45:00Z reading fomo leaderboard, 10 accounts
