Another reminder that this page favors linguistic relativism, for on 4 April 2025 I delivered an illustrative talk where one author issued one task through identical prompts across several languages and the outputs parted ways in register, framing, and risk, with certain responses sliding toward ethnic hostility while others softened claims and shifted agency, so language supplies salience, defaults, and pathways that guide inference and action, and the conduct of large language models renders these pathways starkly visible, which in turn urges a sober revision of the old charge of pseudoscience against Sapir and Whorf.