Innovations in Prediction Markets: Gensyn’s AI-Powered Settlement Approach

‍Investing in digital assets or tokenized equities involves a substantial degree of risk. Market prices can be highly volatile and unpredictable. Movements in the market can result in significant, rapid, and potentially total loss of your invested capital. Before allocating capital, you should carefully evaluate whether such investments align with your financial situation, experience, and risk tolerance. For further information regarding potential risks, please refer to the disclaimer below.

‍

Innovations in Prediction Markets: Gensyn’s AI-Powered Settlement Approach

There are occasional reports of contested market resolutions on prediction markets, often on nuance in the fine print vs. the intention of an event’s question. Participants that believe their position is clearly in the money are surprised when the market resolves otherwise. Ultimately, this issue is around human interpretation of the markets, which can be subject to bias. Gensyn, a decentralized AI infrastructure network, is addressing this by leveraging AI models for event resolutions in prediction markets. This helps deliver predictability and verifiability that a market is settled correctly, ultimately setting the groundwork for agentic trading of these markets.

‍

Gensyn Overview

At its core, Gensyn provides the underlying infrastructure that supports decentralized AI applications. Its Reproducible Execution Environment (REE) toolchain enables consistent machine learning execution across any device, ensuring reproducible results and therefore allowing for the creation of deterministic AI models. In contrast, closed-source AI models can produce different results across multiple runs and different devices due to GPU execution dynamics.  

Gensyn has also developed a trustless method to verify work performed across the network. Paired with its REE, work can be verified without trustless intermediaries. It also has its Agent eXchange Layer (AXL), which allows for P2P communication and data exchanges between machines, supporting use cases including agent-to-agent communication. Finally, Gensyn’s blockchain, a L2 built on the OP Stack, serves as the coordination layer for applications built on it.

The Delphi Information Market

Gensyn launched Delphi, a decentralized information market built on top of its network, in April 2026. Information markets allow anyone to not just trade, but also create markets, differing from prediction markets which have teams dedicated to determining which markets to launch. With Delphi, participants are able to bring their own ideas to the market and can benefit by receiving trading fees for markets they create, something not possible on other prediction markets.

Delphi uses AI models as its ‘judge’ for event resolution, which can limit human bias or manipulation. This has the potential to reduce or eliminate contested settlements. Thanks to its REE, users can verify the AI model’s settlement result on their own, instilling confidence in these markets.

Delphi taps into a rapidly growing market. Prediction markets tracked by data provider Artemis are posting record volume, on pace for nearly $500B of annualized volume based on June figures. However, few if any of these platforms are leveraging AI in this manner.  

How Delphi Works

Delphi is unique in that AI models serve as the judge for all its markets. When creating a market, users have the option to select verifiable and non-verifiable models.

Non-verifiable models are cloud-hosted LLMs such as Claude 4.6. They are able to access data sources on the internet, but may behave non-deterministically, meaning participants must trust that the model and creator handled the settlement process correctly. Without cryptographic proof of the settlement process, there is the potential for surprises on how a market is settled based on a given outcome. These kinds of models can pull the event outcome on their own, thanks to their ability to search data sources on the internet, but users can’t run the same settlement and guarantee an identical result. However, using a top LLM may be seen as less biased for market settlement relative to the status quo.

Verifiable models are locally run and use Gensyn’s REE. When a creator settles a market, they manually retrieve the relevant data from the specified sources, append it to a settlement prompt, and run the prompt through the model inside REE. This produces a settlement receipt, which is a record containing the model used, the full prompt and appended data, and a hash of the computation. Anyone can use the receipt to run the same computation on their own device to verify the settlement was computed correctly.

Verifiable settlement proves that the computation was done correctly given the inputs, but does not on its own confirm where the data provided by the creator is factually accurate. As long as there are clear and identifiable data sources, settlement can be independently audited. 

Nuances of Delphi Markets

Key components of Delphi information markets sit on-chain, such as market creation and resolution. 

On Delphi, a market’s description will show the AI model being used for settlement and whether it is verifiable, the settlement prompt, and the data sources. One consideration with using non-verifiable models is that market setup can be easier – these models are plug and play and can access given data sources directly.

Even with verifiable settlement, resolution can depend on the clarity of the settlement prompt. It is possible that a vague prompt results in a contested settlement result. To help solve for this, Delphi leverages AI tooling to assist participants with market creation. Components such as the question, outcomes, and sources are flagged prior to the markets going live, with suggestions provided.

The settlement deadline is automatically set to 24 hours after the market closes. The market creator must manually settle the market, in part to help deal with unexpected outcomes (such as an event disruption or cancellation). If a market fails to settle within its deadline, the market expires and undergoes an automatic mark-to-market liquidation where participants receive the value of their positions based on the spot price at the time the market closed.

The default trading fee for markets is set to 2%, though creators can adjust it when launching a new market. The creator receives 1.5% of the fees on total traded volume within their market.

Delphi is not registered with the Commodity Futures Trading Commission and may not be available in all jurisdictions.

Takeaways

Delphi is the latest application demonstrating what is possible with Gensyn’s technology. It can enable AI-driven information markets with more predictable market settlement processes, which could be well suited for agentic trading and longer tail markets. Moreover, Gensyn has the potential to expand these kinds of AI oracles to other asset classes and markets.