Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.pelion.dev/llms.txt

Use this file to discover all available pages before exploring further.

This page describes the competitive context. See Status and roadmap for what is currently built.
UMA’s Optimistic Oracle is the incumbent onchain adjudicator. It resolves disputed questions via token-weighted voting among UMA holders. Four failure modes make Pelion’s existence necessary.

Failure mode 1. Whale dominance

Token-weighted voting means a holder’s influence is proportional to their UMA stake. That becomes fatal when two conditions align. The market being resolved has stakes larger than the total market cap of the voting token. A large holder has directional exposure to the underlying market. When both hold, the economically rational move for the whale is to vote against factual accuracy. Honest defense is weaker because honest defenders are smaller and more dispersed.

Case study. The Zelensky-suit market, July 2025

The surface dispute. Did Zelensky wear a suit at the June 2025 NATO summit? Mainstream-media consensus was unambiguous. He did. Multiple photographs, multiple outlets, Reuters and AP both reporting on the record. UMA’s resolution. No. The mechanism that produced that outcome is the point.
MetricValue
Market open interest~$242M
UMA token market cap at resolution~$95M
Ratio (stakes ÷ voting power)~2.5×
The attacker needed to move less UMA weight than the market stakes rewarded. The attack was cheaper than the defense. So the attack succeeded. This is the structural claim. UMA’s economic security scales linearly with its token market cap, while the stakes it adjudicates are unbounded. Every market with stakes greater than UMA’s market cap is a potential exploit waiting for someone to notice and coordinate.

Failure mode 2. Coordination games reward reflexive consensus

UMA’s dispute mechanism penalizes voters whose votes differ from the final resolution. The intent is to punish bad-faith disruption. The effect is to incentivize matching the perceived majority, regardless of factual accuracy. If you believe the resolution will be X for any reason, whether whale pressure, the first announcement, or a high-profile tweet, the expected-value vote is X, because voting against the expected outcome costs you stake. That produces reflexive consensus, where the outcome converges on what voters think will win rather than what actually happened. Reflexive consensus is indistinguishable from truth-tracking when truth is easy. It diverges from truth-tracking the moment a well-resourced actor, or a loud narrative, shifts the perceived majority.

Failure mode 3. Forced binary resolution

UMA must eventually return YES or NO for every market. There is no structural option for “the question cannot be cleanly adjudicated given the available evidence.” Real questions are not always binary. Ambiguous event timing. Did event X happen by deadline Y if the source’s timezone is unclear? Contested primary-source reporting. Two credible outlets report contradictory facts. Edge cases not anticipated by the resolution criteria. UMA collapses all of these into YES or NO. Whichever bucket the ambiguous case falls into, half the market is unhappy and the controversy compounds. Dispute rates correlate with ambiguity, not with attack. Pelion handles this structurally. UNRESOLVABLE is a first-class verdict outcome. The requesting contract decides what to do with it: refund, escalate, retry with refined criteria.

Failure mode 4. Centralization patches don’t address the root cause

UMA’s response to the above has been operational. Managed proposers. Trusted, whitelisted actors propose resolutions to reduce attack surface. Whitelisted resolvers. Certain markets have approved resolvers only. Discretionary escalation. A core team intervenes when disputes escalate. Each of these trades decentralization for quality without fixing the underlying incentive structure. If the voting mechanism produces bad outcomes under adversarial conditions, making voting more restrictive helps. But the cost is that UMA is no longer a permissionless oracle, and the very property that justifies the oracle’s existence (trust-minimized adjudication) erodes.

Why token-weighted voting cannot be patched

The core issue is that token-weighted voting makes economic security proportional to the voting token’s market cap. Any improvement to economic security requires more market cap. More market cap requires buyers. Buyers come because the token captures value. Value capture comes from adjudicating more markets. Adjudicating more markets increases total stakes, which is what created the security gap in the first place. Token-weighted oracles scale their security sub-linearly relative to their workload. Pelion’s insight is that economic security for adjudication should come from a different token economy than the one being adjudicated, and that token economy should be a general-purpose AI work marketplace (Bittensor), not a single-purpose oracle. See Economic security for the quantitative comparison.

What Pelion changes, specifically

PropertyUMAPelion
Economic security anchorUMA token market cap (~$95M)Sum of alpha market caps across routed Bittensor subnets, scaled by quorum requirement
Attack vectorAcquire or coordinate UMAAcquire or coordinate alpha across multiple subnets, plus defeat bonded relay challenge
Failure mode on ambiguityForced binary (YES or NO)UNRESOLVABLE with defined escalation paths
Resolution time (standard)Hours to daysMinutes
Resolution time (contested)DaysHours (challenge window)
Per-query cost~$750 bond threshold for dispute0.01to0.01 to 10 depending on complexity
TransparencyVote totalsFull evidence bundle and reasoning trace per verdict, archived
The last row is the one investors sometimes underweight. Every Pelion verdict ships with its evidence and reasoning, permanently. That is both an accountability mechanism (bad verdicts are traceable) and, over time, a proprietary training dataset for post-event AI adjudication.