Absurd: AI-Native Video Production That's Already Working for Hims, Kalshi, Replit, and Brex

The pitch for AI video has been around for years, but the gap between viral demos and ads that brands will actually run on television, billboards, and paid social has remained stubbornly wide. Absurd, a Y Combinator Fall 2025 company, thinks it has cracked the operating model that closes that gap. The startup describes itself simply as “AI Videos,” but the unglamorous longer version is more interesting: Absurd makes brand and performance ads at scale, combining a multi-agent orchestration layer with a hands-on creative team that works alongside its customers from concept to final cut. Production-quality videos, the company says, can be turned around in 72 hours.

The early traction is the kind that usually takes years to assemble. Companies including Hims, Kalshi, Replit, and Brex have already used Absurd to produce campaigns. The startup’s videos average 400,000-plus organic views, and a recent piece for Kalshi titled “Election Day” cleared a million views – numbers that, in performance-marketing terms, are not just nice-to-haves but the difference between a creative experiment and a real distribution channel. In a world where every brand is racing to feed an insatiable demand for short-form video, the ability to ship AI video ads for brands in days rather than months is a meaningful unlock.

Absurd was founded in 2025 by Philip Ho (Founder and CEO), Damian Chng (Founder), and Daniel Kathein (Founder and CTO), a six-person team that draws on prior experience at companies like Meta and NVIDIA. The combination – a deeply technical core paired with operators who understand what brands actually need to ship paid creative – is well suited to a market that rewards speed, taste, and reliability in equal measure. With AI now firmly embedded in the broader generative AI, video, media, marketing, and advertising stack, Absurd is positioning itself as the production partner that lets brands finally treat AI video as a primary channel rather than an experimental side project.