[ethan_smith]

These percentage estimates of AI's proximity to "human reasoning" are misleading abstractions that mask fundamental qualitative differences in how LLMs and humans process information.