Whatnot bought Shaped because live auctions don't give recommendation software the luxury of being late.
The livestream shopping platform has acquired Shaped, a machine learning startup that builds real-time recommendation and search infrastructure, according to a TechCrunch report published July 15, 2026. The deal brings Shaped founder and CEO Tullie Murrell, a former Meta engineer, into Whatnot along with roughly a dozen engineers and researchers. Murrell will run a new Applied AI Research group inside the company.
This is not a cosmetic AI hire. Whatnot's whole product depends on showing you the right stream, seller, item and auction before you drift away. In a live marketplace, that window can be measured in seconds.
Why Shaped Fits Whatnot
Shaped built its business on a narrow but valuable idea: use a company's own customer data alongside large language models to make search and recommendation results change as a session unfolds, not after a daily batch job. QVC used the system for real-time video recommendations inside its mobile app. Outdoorsy, the RV and campervan rental marketplace, used it to improve search relevance from live session activity.
Two very different businesses had the same problem. A browsing customer leaves quickly when the next result feels stale.
That problem is sharper at Whatnot than it is in ordinary retail. A live auction doesn't sit still. Inventory disappears when a seller taps sold. Prices move as bidders jump in. A recommendation engine built for a static product catalog is already behind by the time it renders. Old data, no good.
TechCrunch reported that Whatnot has spent six years cutting the time it takes to generate a personalized recommendation from roughly a day to minutes. Folding in Shaped is meant to close more of that gap. You don't buy a team like this to make the homepage a little nicer. You buy it because latency is now part of the product.
Buying The Team Says Plenty
Whatnot isn't short of cash. The company raised a $225 million Series F in 2025, co-led by DST Global and CapitalG, with Sequoia Capital and Alkeon Capital joining as new investors, according to Crunchbase News. That round more than doubled its valuation to $11.5 billion in under ten months. Fortune also reported in June 2025 that sellers on the platform had crossed one billion cumulative orders.
It bought instead.
That choice says something. Recommendation and search infrastructure for real-time, high-churn markets is hard to hire for and slow to build from scratch. Whatnot apparently decided a working team with deployments at QVC and Outdoorsy was worth more than another year of internal engineering sprints.
Frankly, the acquisition price is the less interesting part, and neither company disclosed it anyway. The better question is what live commerce is turning into as a testing ground. Static e-commerce recommendation, the kind Amazon and Netflix refined over two decades, can tolerate delay. You can recompute a homepage overnight. A live auction cannot wait.
That is the bet. If Whatnot can personalize shopping under the pressure of vanishing inventory, moving bids and short attention spans, the lessons may travel back into less frantic parts of retail. The hard version teaches you faster.
Shaped raised an $8 million Series A in 2024 and built its self-serve platform around exactly that pitch: recommendations that respond to what a shopper is doing right now, not only what they did last week. It was still a small company. But landing customers such as QVC gave it proof that the software could run outside a demo and inside real consumer apps.
There is still one unresolved detail. What happens to Shaped's existing clients now that its core team sits inside Whatnot? Neither company has said whether QVC and Outdoorsy will keep using the technology. For Whatnot, the logic is clear enough. If live shopping is going to be more than people shouting over fast-moving auctions, the platform has to know what you want before the room moves on.
Also read: Bitcoin Miner MARA Buys Texas Power Site to Chase the AI Boom • A Caltech Startup Shrank a 27 Billion Parameter AI Model to Fit on an iPhone • Anthropic Bets the Next Trillion Dollar Business Is Installing AI, Not Building It
