AI is moving from a tool people operate to a system that acts on their behalf. That changes how products are built, software is distributed, and which human skills matter.
OpenAI shared a useful framework for spotting assumptions that may expire:
-
Agents will perform a growing share of Internet actions. Agent traffic changes software economics, not just interfaces. What happens when software’s primary operator changes?
-
Assume models will be substantially better in a year. What becomes possible as today’s constraints disappear?




