For years, quant finance hid behind proprietary models and private backtests. That era is ending.

In this piece, we explore why open-source benchmarks and agentic research systems are quietly exposing the real constraints of alpha generation. The result is that many sophisticated AI models fail once evaluation becomes realistic. What survives is not complexity, but judgment, alignment, and the ability to adapt as markets change.