UID 221 Claims Crown; Eval Schema Updated

UID 221 dethroned the previous king with a composite score of 0.493, earning ~26.3 τ/day from the crown pool. The team pushed an eval update (commit 4c54c83) that improves chat-shaped probes to honor thinking mode, raises code/reasoning/tool-use sample counts, and reweights the composite scoring to emphasize tool use, chat quality, and code over raw KL. No hard per-axis king gates were introduced; changes are soft incentives via ranking composites.

• •UID 221 dethroned UID 82 with 0.493 composite (0.391 prior).
• •Chat probes now support thinking mode; thinking-collapse enabled by default.
• •Code, reasoning, tool-use samples increased; composite weighting adjusted.
• •No hard per-axis king-quality floors; soft incentives only.