gfermah A prediction market can attract attention and still face a problem at settlement. What exactly counts as evidence that an event happened? In May 2026, Strategy sold 32 Bitcoin between May 26 and May 31, but disclosed the sale in an SEC filing on June 1. A Polymarket market asking whether Strategy would sell Bitcoin by May 31 became disputed because traders disagreed over whether the outcome should follow the transaction date or the date the sale became public. This resolution problem makes Ultramint @ultramint_xyz worth examining. It is the first application built on Fermah @fermah_xyz Kernel, and its design puts data sources, resolution rules and dispute handling at the centre of each market. Now, let's explain it better 👇 Define the evidence before trading begins Ultramint's Terms describe a market setup where a designated data source and resolution rule are selected when the market is created and disclosed before users take positions. That distinction matters. A question about whether a company sold Bitcoin by a deadline must specify whether it measures the transaction date or public disclosure date. In the Strategy case, the 8-K filing reported that the 32 BTC sale took place during May 26 to May 31. The filing appeared on June 1. Polymarket's original rules named company information and onchain data as primary sources, with credible reporting as backup. But the disagreement centred on whether confirmation also had to be public before the deadline. A clear source and rule give participants a defined basis for evaluating the outcome before money enters the market. Turn the rule into a checkable outcome Ultramint documents a deterministic Rule Engine. It applies the market's pre-defined resolution rule to the output of its designated data source. For structured questions, its source layer can use services such as Pyth price feeds, sports data APIs and social platform APIs. The engine evaluates the data against the stated condition rather than introducing a new interpretation after the event. For open-domain questions without one authoritative data field, Ultramint describes as an AI Oracle. It retrieves and scores evidence from a moderated trusted-domain allowlist, then submits that evidence to independent language-model judges operating under quorum, supermajority and model-diversity requirements. A market about a published price can use a structured feed. A question requiring evidence from several trusted sources can use the AI Oracle process. Give disputed outcomes a defined path For automated markets configured to use it, Ultramint describes a proposal, bonded-challenge, arbiter and finalization sequence. A proposed result can be challenged by posting a bond, after which an arbiter set determines the dispute. This protection does not apply to every market. Ultramint's Terms also describe creator-resolved markets without the same challenge process. The resolver type and its protections matter before anyone participates. In the Strategy case, the market resolved No after repeated challenges and a final UMA vote backed by 98.6% of voting power. The sale had occurred before the deadline, but public confirmation arrived after it. The dispute shows why a market must define in advance how those two dates affect settlement. Why this matters for Kernel The point is not that Ultramint would automatically have resolved the Polymarket dispute differently. Its design makes source selection, resolution rules and dispute handling explicit parts of market creation. Through Ultramint, Fermah Kernel is being used to coordinate how a market defines evidence, evaluates outcomes and reaches settlement. Prediction markets need more than participation and liquidity. They need rules that remain clear when the real world gets complicated. @fermah_xyz @ultramint_xyz @7wealthh