A similar perspective as mine: https://ianreppel.org/the-ai-benchmark-trap/