The problem Polymaxx.health solves

Polymaxx.health is a real-time market intelligence platform that applies 0G’s GLM via a claude skill to high-frequency predictive markets such as weather forecasting and TSA passenger volume projections.

Clients, referred to as consumers on our platform, subscribe to a WebSocket feed that delivers live signals identifying potential market arbitrage opportunities. When a signal is broadcast, consumers compete in a rapid bidding process to gain access to the full trade signal. Bidding occurs within seconds, and consumers who fail to respond in time are automatically excluded from that round.

Each consumer runs a signal evaluation agent powered by 0G Compute for dynamic pricing analysis. This agent determines whether the opportunity meets predefined profitability criteria. If the signal qualifies, the agent notifies Polymaxx to execute the trade.

On the backend, we use the x402 protocol to handle payment submission for access to the full signal. By integrating Kite AI along with a reference x402 implementation, we support decentralized identities (DIDs) and on-chain reputation scoring. This ensures that consumers who place bids they cannot fulfill are penalized or removed, maintaining marketplace integrity.

Challenges we ran into

We were unable to obtain API keys for Kite AI during the hackathon. As a result, we implemented the integration based on the available documentation and built a reference vanilla x402 payment flow to ensure there was a working, end-to-end implementation.

Collecting high-quality market signals required significant time and effort. In addition to ingestion, we had to design reliable search and indexing pipelines so the signals could be surfaced accurately within our dashboards.

To validate the system, we executed live trades on Polymarket. Because these were real-money transactions, we had to be deliberate and selective in our testing strategy to manage risk and control costs while still validating the full execution flow.

Use of AI tools and agents

We used Claude Code Opus 4.6 extensively to create dashboards, do research on how to use tool kits from 0g and Kite API, as well as getting Polymarket testing going.

We also used Cursor with Context7, Tavily, and Github MCPs.