Shaped.

Every time a feed reshuffles, a search box guesses your intent, or a shopping app pushes the one item you didn't know you wanted, something is quietly ranking millions of options and picking a handful. For a growing roster of marketplaces, media brands, and stores, that something was Shaped - a New York company that treated ranking not as a feature but as infrastructure.

By the YesPress Desk · Filed July 2026

Named customers spanned prediction markets and camper-van rentals alike - Vox, New York Magazine, Outdoorsy, QVC, Kogan, and Polymarket among them.

Recommendations used to require a data-science team and six months. Shaped turned it into an API call - changing who its customer could be.

Shaped sat at the intersection of three markets converging at once: recommendation systems, enterprise search, and the retrieval layer powering the new wave of AI agents. Its 2025 repositioning - from "recommendations and search" to "the real-time context engine for agentic AI" - was a bet that memory and retrieval are, at heart, ranking problems with a clock attached.

The acquisition was itself a market signal. Whatnot, a fast-growing live-commerce marketplace, chose to buy the team rather than build the capability - a sign that real-time discovery infrastructure has become a build-vs-buy decision at the largest platforms. Murrell now leads Whatnot's Applied AI Research group, with roughly a dozen Shaped engineers alongside him.