The model's own turn, replayed into history so the results that follow have something to answer.
Its reasoning blocks travel with it: Anthropic asks for a tool-use turn's thinking to come back
unchanged, and the driver replays them ahead of the calls.
The model's own turn, replayed into history so the results that follow have something to answer. Its reasoning blocks travel with it: Anthropic asks for a tool-use turn's thinking to come back unchanged, and the driver replays them ahead of the calls.