Updated
Updated · TechCrunch · Jul 15
Whatnot Buys Shaped to Sharpen AI Recommendations as 500,000 Weekly Live-Video Hours Strain Discovery
Updated
Updated · TechCrunch · Jul 15

Whatnot Buys Shaped to Sharpen AI Recommendations as 500,000 Weekly Live-Video Hours Strain Discovery

1 articles · Updated · TechCrunch · Jul 15

Summary

  • Whatnot said Shaped’s recommendation and search technology will push product discovery closer to real time on a marketplace where auctions, inventory and buyer demand shift by the second.
  • Six years of in-house work cut recommendation latency from about a day to minutes, and the deal is meant to speed that further across more than 500,000 hours of live video and millions of weekly interactions.
  • Nearly a dozen Shaped engineers and AI researchers will join Whatnot, with founder Tullie Murrell set to lead a newly created Applied AI Research group.
  • The acquisition lands as Whatnot scales rapidly: sellers have topped 1 billion orders, the company added 20 million buyers in the past year, and a $225 million Series F valued it above $11 billion.
  • That expansion has widened the recommendation challenge as Whatnot launched 35 new categories last year and 45 more in the first half of 2025, while rivals including eBay and Poshmark also deepen AI efforts.

Insights

Has Whatnot's AI solved its discovery problem or created new challenges for its millions of sellers?
Can Whatnot’s AI engine defend its turf from e-commerce giants like Amazon and TikTok?