If you’ve consistently used the frontier models over the past year and a half, it’s hard not to be astonished by their trajectory. They are very intelligent. As he says, 2025 for agents in the work force seems reasonable. Brilliant ai co-workers at scale seems plausible if we adapt to use them.