Why shopping apps are buying AI startups

The livestream shopping platform Whatnot has acquired Shaped, a startup that specializes in building recommendation systems. Unlike a standard online store where items sit in a inventory for months, Whatnot hosts thousands of live auctions that start and end constantly. This creates a moving target for software trying to suggest products. By bringing Shaped’s team in-house, Whatnot aims to sharpen its ability to process millions of user interactions each week. The goal is to evolve its recommendation engine—the behind-the-scenes software that suggests what you might like—to become faster and more responsive to the chaotic, split-second nature of live commerce.

Recommendation systems rely on machine learning, a method of teaching software to recognize patterns in data. Think of it like a store clerk who has a photographic memory of every item passing through the shop and every customer's past purchases. As you watch a stream, the system continuously tracks what you click, how long you stay, and what you ignore. It uses this stream of data to build a personalized profile of your taste. While large language models—which are systems trained to process and predict structures in text—are used here to help classify products and improve how discovery works, the real-time heavy lifting is done by specialized algorithms that match your profile against active auctions. These algorithms ensure that when one auction ends and another starts, your screen immediately updates with new options that belong in your specific neighborhood of interest.

We are entering a future where our digital experiences are entirely bespoke. In a traditional physical store, the shelves stay put. In a personalized app, the store layout is essentially being rebuilt for you every time you refresh the page. While this makes it easier to track down niche interests like vintage golf gear or rare art, it also highlights how much of our shopping is now guided by unseen, constantly-updating math. As platforms race to adopt these tools, the way we browse is shifting from active searching to a passive experience where we simply watch what the AI suggests. The companies that win will be the ones that can process your shifting intentions fast enough to keep up with the broadcast.