What's this about?
Mollick posts a side-by-side: "10 months between these two visions of how ChatGPT agents work."
Why is ten months the right unit?
Because it is short enough that the earlier vision is still in living memory, and long enough that the change is total. The pace makes conventional intuitions about technology adoption useless — the mental model you formed by trying something last autumn describes a product that no longer exists. This is the concrete illustration of his ability-perception gap argument from two days ago: perception updates on the timescale of personal experience, and capability moves faster than people re-sample.
Why is it worth noticing?
235 likes
- It is a visual argument, which is why it travels.
- It is also a caution about this digest and every other: anything written about agent capability has a short shelf life.
- Ten months ago is roughly when "agent" became a product category rather than a research term.