GPT-5.4 Scored Higher Than Humans on Productivity Tests. That Should Make Us Uncomfortable

A model just completed office work more reliably than the humans used as the benchmark. We noted it and just moved on.

Continue reading on Medium »