Why Does AI Need a New Internet?

Parag Agrawal says AI needs a new internet because agents will use the web on a scale today's systems were not built for, and the bargain that keeps the web open fails if it does not change. About two and a half years before this conversation he wrote down a figure: a thousand times more use than humans have ever made of the web. He had built systems at that multiple before, and he says none of them were designed for it, in architecture or in infrastructure. Business models evolved for human distribution and monetization, over what he calls about three decades, have to change with that use.

He calls the shift a second birth of the open web, a parallel web. The name came later, not at the founding moment. People publish in the open because distribution and money reward it. If agents read that content and the incentives are not rebuilt, the web starts closing, and he says that closing has already started. He thinks a thousand times is already too low. His own heavy token use of the web is at that multiple, he expects the real multiple to go higher, and he thinks the use creates real value rather than pure waste.

The mission he states for Parallel is to keep the web open by aligning that value with the people who curate and produce the information. Reward high-quality, distinct content, and the open web can become something new.