GLM and Muse completed; DeepSeek stopped after 10 behavior-readable episodes.
See analysis, design, series protocol, and the GLM, Muse, and DeepSeek plans.