When agents operate the apps, who owns the customer relationship?

Thompson starts with a small experiment: an agent organized his saved Instagram recipes into a personal app. It took minutes, but extracting the actual recipes remained unfinished and constrained by Instagram’s rate limits.

His larger argument is that useful agents need computers of their own, including persistent cloud environments. They can then operate services and generate interfaces for an individual user, reducing the need to navigate each supplier’s app.

Aggregation Theory originally explained how abundant online content made discovery—and ownership of the audience—valuable. Thompson proposes a further shift: when executing tasks becomes abundant, deciding what to do becomes scarcer. The agent that inspires and carries out an intention could control demand, leaving underlying apps competing to fulfil it.

This is a strategic thesis, not evidence that the transition is complete. His own recipe example exposes continuing dependence on access, extraction and platform restrictions. The compelling question is less whether apps disappear than whether customers still choose them directly, or increasingly accept an agent’s selection.