Mantle Launches Turing Test Hackathon to Benchmark On-Chain AI

Mantle announced the Turing Test Hackathon 2026, a two-phase competition with a $120,000 total prize pool that asks teams to deploy autonomous agents on Mantle and compete across trading, data, and infrastructure challenges. The initiative is co-hosted or backed by major ecosystem players including Bybit, Byreal, and Blockchain for Good Alliance (BGA), and sponsored by Tencent Cloud, with additional partners such as Mirana Ventures and others. Phase 1, ClawHack, launched April 15 with $20,000 in prizes and runs through April 30. Phase 2, AI Awakening, carries $100,000 and introduces a Human vs. AI mechanism across six competitive tracks.

The hackathon requires participants to run agents that interact natively with Mantle's on-chain DeFi primitives. Organizers emphasize permanent, auditable recording: every decision and outcome will be logged on-chain, creating immutable traces for evaluation. Byreal supplies an AI agent trading platform integrated with Mantle; participants can use provided Python and Solidity templates and connect to exchange liquidity via Bybit APIs. Evaluation criteria mentioned include trading volume, return on investment, on-chain anomaly detection, and smart-money tracking performance. The event also lists ecosystem tooling and data partners such as Nansen, OraKle, and cloud infrastructure support from Tencent Cloud, which suggests participants will have access to off-chain compute, real-time oracles, and enriched on-chain analytics for agent decision-making.

Sponsors and partners span trading platforms, analytics firms, accelerators, and developer communities, which signals a push to blend AI agent research with real-world financial flows. Community and academic partners include groups from Cornell, Berkeley, HKUST, and numerous regional blockchain communities. The competition explicitly supports agentic AI use cases such as AI-driven trading, on-chain anomaly detection, and data-driven alpha discovery, with a Human vs. AI track designed to compare human-crafted strategies against fully autonomous agents.

This hackathon frames itself as the first large-scale benchmark that uses an on-chain environment to assess agentic AI performance at scale. For practitioners that build agentic systems, the combination of immutable logging, economic incentives, and live-market interactions creates a reproducible, adversarial environment for testing robustness, latency tolerance, and economic safety. The event also highlights convergence between Web3 primitives and autonomous AI: on-chain transparency improves auditability but introduces new operational risks, including front-running, oracle manipulation, and exposure of proprietary strategy logic. The inclusion of cloud providers and analytics partners underscores that hybrid on-chain/off-chain stacks remain central to production-grade agent deployments.

Track submissions and published on-chain traces to see whether Mantle releases a standardized benchmark or dataset for agent evaluation. Monitor how organizers mitigate economic abuse vectors and whether winning strategies are reproducible, sanitizable for publication, or monetized. Watch for integrations from exchanges and oracle providers that could become de facto tooling for future agentic systems.