Ex-Ramp engineers raise $25 million for creative AI agents after scrapping their first idea

CRV put up four of every five dollars in the raise, and its Series A is four times the size of the seed ^[16]. Both checks went to a company founded in 2026 ^[17]. Its founders all worked as engineers at Ramp and first set out to build an AI-powered performance marketing tool ^[4]. After six months they decided that idea did not "have legs," Joowon Kim told TechCrunch, according to PYMNTS, and they turned to tools for making creative assets and campaigns ^[10]. "We scrapped the entire codebase; we burned all of it," he said ^[11].

If those six months fell inside 2026, the creative platform was no more than about three months old when the funding was announced on Oct. 6 ^[1]. Oct. 6 is the 279th day of the year. Take away roughly 182 days for the abandoned idea and about 97 remain ^[18].

The case for agents over conventional creative software comes from Kim. "Transitioning from wanting to make something to actually making it has never been seamless," he wrote in the announcement ^[7]. Users, he added, should be able to say things like "expose a more natural lighting" and let "agents obsess over the menial tasks that never defined what it meant to be creative" ^[13]. The announcement, as reported, does not include a valuation, revenue or a customer count. For now, the claim that agents take production work away from design software is Kim's argument.

One reading is that General Catalyst and CRV ^[5] paid for founders who will kill an idea at six months. If so, a second pivot would still fit what they signed up for. Another reading is that the pivot moved less than the burned codebase suggests. The platform is built for images, videos and ad campaigns ^[2], and I'd expect the people who buy ad campaigns to overlap heavily with the people who would have bought a performance marketing tool. A third is a consumer business, where payers are rare but spend a lot: the top 1% of consumer AI spenders averaged $903 a month in August, an 80% increase over the prior 18 months, according to an Andreessen Horowitz analysis of YipitData figures ^[14].

I think the second reading is the likeliest. A startup building out a sales organization ^[6] is usually selling to companies. The counter-thesis is Kim's own: a product with no learning curve can sell to anyone with an idea. A next round priced on revenue from customers who cut design-software spending in favor of Melius agents would prove him right and this column too cautious.