Pay Publishers Per Inference
Parag Agrawal, former Twitter CEO, launches Index at Parallel Web Systems. The platform pays publishers and creators when AI agents use their work, calculated by Shapley value at inference time. Parallel hit $2B valuation in April.
First credible attempt to price content for agents, not humans. Shapley value scores each source's contribution to the agent's output at inference time. Not per crawl. Not per citation. Per slice of agent work.
Launch partners are who you'd want: The Atlantic, Fortune, Every, Not Boring, The Generalist, Exponential View. Data providers too: PitchBook, ZoomInfo, Tracxn. High-value content for high-value agent tasks pays more. Two weeks later, Ben Thompson interviewed Agrawal on Stratechery about agentic content economics.
For PMs building agentic products: assume content costs become a real line item by 2027. For content businesses: the agentic-traffic monetization layer just emerged. For execs: ad-supported web is shrinking. Agent traffic doesn't see ads.