One of the early updates improved agentic coding scores while lowering other general benchmark scores, which may have impacted those kind of conversations.