Dolores Research has launched WAGMI Bench, an open-source framework designed to provide standardized, reproducible testing for AI trading agents, modeled on SWE-bench’s role in evaluating coding agents. The benchmark covers 13 historical periods in the BTC perpetuals market, from the COVID crash to the period surrounding ETH approvals, with each agent assessed at 3,150 decision points under consistent conditions. Rather than emphasizing profit and loss, the framework focuses on survival and engagement metrics and does not permit favorable time-window selection or omission of drawdowns. The first Classic 13 study, version 1, tested five models and one mechanical baseline without declaring a winner. Released under an Apache-2.0 license, WAGMI Bench can be forked, extended and used by developers to test their own agents. Dolores Research also launched the $DOLORES community token through Virtuals Protocol on Robinhood Chain, allocating 2% of its total supply to veVIRTUAL stakers. The token is intended as a coordination and research-support layer, not a governance mechanism, and does not affect benchmark results or agent performance metrics. The launch came amid a broader promotional wave around Virtuals Protocol in late August 2026. The framework arrives as agent-based trading infrastructure expands, following the CFTC (U.S. derivatives regulator) approval of a true BTC perpetual in May 2026 and infrastructure investments by Coinbase and Binance aimed at enabling automated systems to interact with their platforms. WAGMI Bench gives developers a common set of conditions for apples-to-apples comparisons.