There are two key issues in prediction markets:

The First Challenge: Market creation and resolution

Not all topics can be turned into prediction markets—viable topics need a certain level of public interest. Topics that are too niche may attract little attention or be prone to manipulation. Additionally, prediction markets require clear rules to define how they resolve—rules must be unambiguous and backed by publicly verifiable sources. Polymarket, for example, has seen cases where vague rules led to disputes among users over market outcomes.

These challenges fall under the domain of NLP, an area where AI has a strong advantage. AI can effectively filter spam and generate precise market rules. With the LLMs, the cost of developing in this field has significantly decreased. Buzzing leverages RAG on top of LLMs to build a dedicated prediction market agent. Curators only need to input a news headline, and Buzzing’s agent will search the web for relevant information, assess the topic’s quality, and generate rigorous market rules. We have completed the development of a highly effective AI agent.
Additionally, we use AI agents for market resolution, with UMA’s oracle serving as a fallback to handle edge cases where the AI agent makes errors. UMA’s oracle still requires human intervention, and for a permissionless prediction market with potentially millions of topics, resolving every market through UMA alone would be expensive. Using AI agents helps reduce costs and achieve scalability. Once Buzzing launches its token, we also plan to introduce DAO governance as a fallback mechanism for AI agents in market resolution.

We used another scoring AI agent to evaluate the prediction market’s AI agent, with an average score of about 9.5.

The Second Challenge: Liquidity

Liquidity has always been a major challenge in prediction markets. The core issue is that DeFi whales, who typically provide liquidity, often lack knowledge about prediction market topics and have no clear expectations of returns or risks. On the other hand, curators, who understand prediction markets well, usually lack the capital to provide liquidity. To address this, Buzzing has designed the BLP vault to supply bootstrapping liquidity for new markets and supplement liquidity for high-demand markets:

For DeFi whales, the USD vault provides liquidity to all markets, eliminating the complexity of allocating liquidity to individual markets. Additionally, since the BLP vault acts as the counterparty to all traders, it is inherently profitable—because traders, as a whole, tend to lose money over time.

For curators, BLP’s bootstrapping liquidity functions similarly to TikTok boost new videos. It removes liquidity concerns for curators, allowing them to focus on market creation and promotion. When curators create a new market, they must deposit a small amount of funds into the BLP vault. This mechanism aligns curators’ incentives with BLP, preventing Sybil attacks in low-value markets

We have completed the development of the BLP vault and conducted backtesting using historical transactions from Polymarket. The results show that the vault can meet trading demand while achieving better profitability, especially in markets with lower trading volume.