Pandora is a decentralized prediction market protocol built on the Ethereum blockchain. The platform allows users to create markets and trade on the outcomes of any real-world event.
It is distinguished by its use of an artificial intelligence (AI)-powered oracle system for resolving market outcomes and a dual-engine architecture that supports both parimutuel and Automated Market Maker (AMM) trading models. [1] [2]
Pandora launched on the Ethereum mainnet on February 23, 2026. The protocol was developed by an AI-focused team known as Hey Anon, led by Daniele Sestagalli. The stated mission of the project is to provide a transparent and decentralized alternative to what the creators describe as the "prediction market duopoly," criticizing incumbent platforms for operating with opaque, "black box" market-making practices. [2] [5]
The core of the platform is its permissionless nature, which allows any user to create a prediction market on topics ranging from global politics to internet culture. All market mechanics are designed to be fully on-chain, ensuring public verifiability. Market resolution is handled by the Pandora Oracle, a system that relies on a consensus of multiple proprietary AI agents.
In cases where the AI cannot reach a consensus, the decision is escalated to a vote by stakers of the $ANON token. The platform's long-term vision is to become a "home for trading agents," enabling both human and autonomous AI to interact with the protocol. Future plans include allowing AI agents to autonomously assess markets, optimize strategies, and deploy new markets, with a scoring mechanism to rate agents based on performance and integrity. [1] [2] [5]
As of March 3, 2026, shortly after its launch, the platform had recorded a total trading volume of $167,900 across 246 created markets, with 1,107 total trades executed. [1]
Pandora was officially launched on the Ethereum network on February 23, 2026. The launch was announced by its development entity, Hey Anon, through a press release. The initial announcement highlighted the platform's AI-driven oracle and its aim to bring greater transparency to the prediction market sector. [1] [2]
Pandora integrates several key technologies to operate its decentralized prediction market, focusing on a transparent on-chain architecture, flexible trading models, and an AI-based resolution system. [2]
The protocol is built on the Ethereum Layer 1 blockchain. A core design principle is transparency, with all market mechanics, trades, and fund management executed via publicly verifiable smart contracts. The protocol operates on a non-custodial basis, which means users retain full control and custody of their own funds without depositing them into a centrally controlled entity. [1] [2]
The protocol’s technical foundation consists of several key smart contracts. A PredictionOracle contract manages the creation and resolution of "Polls"—the underlying questions with defined rules and resolution sources. A MarketFactory contract then creates tradeable "Markets" (either AMM or Pari-Mutuel) that are linked to these polls. The system also uses a 5-minute time window, or "Epoch," for managing deadlines. [6]
While the protocol is described as non-custodial, the available source materials do not mention if the smart contracts have undergone any third-party security audits. [1]
Pandora employs a "dual-engine architecture" that supports two different types of trading mechanics, allowing market creators to choose the model best suited for their prediction. [1]
This model is designed for binary (Yes/No) outcome markets. In a parimutuel system, all bets are placed into a pool. The odds are dynamic and change as more bets are placed on one outcome versus the other. After the event is resolved, the total pool (minus protocol fees) is distributed among those who bet on the correct outcome. This model can enable high-leverage betting opportunities. [1] [6]
This model, also referred to as V2, is used for markets that may have multiple possible outcomes and facilitates continuous trading. It utilizes a Uniswap-style AMM pricing model for YES/NO outcome tokens to provide instant liquidity, allowing traders to buy or sell outcome shares at any time without needing a direct counterparty. This facilitates ongoing price discovery up until the market's resolution. [1] [6]
A key feature of the protocol is its unique oracle system, a smart contract named PredictionOracle, for resolving market outcomes. Instead of relying on a centralized entity or a traditional, human-based oracle network, Pandora uses a proprietary, AI-driven process. [1] [6]
The resolution process follows two main steps:
$ANON Stakers: If the AI agents are unable to reach a consensus, the resolution task is escalated to the stakers of the $ANON token. These token holders then vote to determine the final, correct outcome. [2]This system was designed to minimize human interference, reduce the potential for manual resolution errors, and increase the overall decentralization of the platform. [2]
Pandora offers several user-facing features to support its trading environment.
Hey Anon is the development team and decentralized finance (DeFi) protocol behind Pandora. The group is led by Daniele Sestagalli. Hey Anon focuses on applying AI to simplify complex activities within the DeFi space, using a combination of natural language interfaces and machine intelligence for tasks like portfolio management and the execution of automated strategies. [4] [3]
Pandora represents one of its flagship products, applying its AI focus to the prediction market sector. [2]
The $ANON token is the native asset associated with the Hey Anon ecosystem and serves a critical function within the Pandora protocol. Its primary utility is related to governance and security of the oracle system. Stakers of the $ANON token act as the final arbiters for market resolution.
They are called upon to vote on an outcome if and only if Pandora's primary AI oracle system fails to reach a consensus. The provided source material does not specify other tokenomics, such as supply, distribution, or other potential use cases. [2]