Judgment Is the Bottleneck in Crypto Some questions need someone to read the evidence, not just a price feed. Did an event happen before a deadline? Does a bug report meet the stated criteria? Is a flagged account actually fraudulent ? These questions depend on language, timing, and evidence. Here are five examples, and where a system like @GenLayer could help. Polymarket: resolving a market @Polymarket resolves markets through UMA’s optimistic oracle. A proposed outcome is backed by a bond of about $750. Others have two hours to challenge it. If someone does, UMA token holders vote on the result, a process that can take four to six days. In the Strategy bitcoin market, the dispute turns on timing. Between May 26 and May 31, Strategy sold 32 BTC, roughly $2.5 million. The sale was disclosed in an 8-K filing on June 1, after the market’s May 31 deadline. The market resolved “No” twice, and both outcomes were challenged. A system could compare the market’s wording with public filings and trading records, then return a decision with reasons. This is a possible use case, not an existing connection: none of the sources reviewed links Polymarket to GenLayer. Hyperliquid: deciding which addresses qualify for settlement In March 2025, a whale’s manipulation of the JELLY token left the community-owned HLP vault with roughly $12 million in unrealized losses at first. @HyperliquidX 's validators voted to delist the JELLY perps. The settlement covered all users except addresses that had been flagged. The rule used to flag those addresses was not public in the sources reviewed. Publishing the criteria, and explaining why each address was excluded, would make decisions like this easier to review. A system could apply those criteria to on-chain evidence, provided the rules and evidence were accessible. Immunefi: grading bug reports @immunefi grades bugs by impact. Its published criteria include examples, such as direct theft of user funds rated Critical. They also say a bug may be downgraded if the attacker needs privileges. The page does not describe an appeal process. A system could compare a report with the published rubric and identify which criteria support the assigned severity. But reports can contain details that enable an exploit, so access and disclosure risk matter. High-severity cases still need human review. Optimism: reviewing project funding @Optimism ’s Round 3 approved more than 600 projects. Reviewing that many proposals puts a heavy load on voters. Some teams also split their work into smaller contributions to earn more. Applying a published, versioned rubric to public evidence could make reviews more consistent and give a written reason for each score. But a system could also be gamed, and this use case has not been tested. Gitcoin: reviewing Sybil flags In @gitcoin 's GR14 round in 2022, 35.8% of users were flagged. The brief’s own estimate of Sybil incidence was 22.1% (±3%). The process recorded about 42 appeals and 40 grant disputes. Of the appeals, 22 were denied, 10 were approved, and 10 moved to other round hosts. The cost per grant review was $3.30. The brief reports a flagging efficiency of 162.1%, which it reads as over-flagging. A system could help review appeals against the available evidence. It could not, by itself, correct bias in the detector that flagged the accounts. What these cases have in common Each case depends on text, timing, or evidence, not only on a value a price feed can report. Today, those decisions involve token-holder votes, human reviewers, validators, or automated detection. GenLayer’s Intelligent Contracts are designed for outcomes that depend on natural language and live web data. A validator committee checks a proposed result against the contract’s equivalence rule. One leader proposes a result, the other validators check it independently, and the result is accepted only if a majority agrees. This can provide a way to build evidence-based decisions on-chain. It does not make a decision correct just because validators agree. Limits The criteria must be clear. The evidence must be available and reliable. Validators must check the result, not simply accept the first proposal. A system cannot fix biased source data or a flawed detection tool on its own. None of the sources reviewed links these five projects to GenLayer. The whitepaper figures are design defaults, and GenLayer’s fees and finality times have not been published. @kstellana @RuzgarFlns @Aezakmi_x