rollout and correctness review
Tool Deferral With Essential Capabilities Preserved — PR #97979
Reviewed the core-tool deferral rollout against real capability use, schema validation, and compatibility seams. I called for structured clarification to remain immediately visible after Teknium's A/B evaluation showed a regression; the merged change keeps it outside the default deferred set. The implementation and evaluation are Teknium's work.
Reviewed the core-tool deferral rollout against real capability use, schema validation, and compatibility seams. I called for structured clarification to remain immediately visible after Teknium's A/B evaluation showed a regression; the merged change keeps it outside the default deferred set. The implementation and evaluation are Teknium's work.
Teknium conducted the A/B discovery and implemented the change. My contribution was the documented review request.
