Sometimes you’ll be able to only discover what’s fallacious after you’ve got been involved in the process for a while. Such was the case with Michelle Lim, who, while marketing for Warp’s growth throughout the past 12 months, realized that the company wasn’t updating its website quickly enough.
She noticed that potential customers were asking ChatGPT and other AI bots every kind of questions about Warp’s offerings, but the information they were looking for, comparable to how the product compared to a newer competitor, wasn’t available on the startup’s website. Lim believed this content gap would develop into much more critical as next-generation artificial intelligence agents begin actively crawling the Internet to gather information for users.
It was clear that Warp needed to add more content, but creating and uploading each additional website was a time-consuming task, involving the design agency and many people from different departments.
“Marketers can’t wait a month for design and development teams to build a site,” she told TechCrunch. “With AI engines, you need to create content much faster than before to meet consumer demand.”
Lim, who had been planning to start a startup for a very long time, realized that it was quickly becoming a problem that needed solving. So in March she co-founded Flintan AI platform that enables you to arrange websites that update themselves. She was joined in this effort by Max Levenson, an engineer who previously led the simulation and infrastructure teams at autonomous vehicle startup Nuro.
On Tuesday, Flint emerged from stealth mode with $5 million in seed capital. The investment was led by Accel and included participation from Sheryl Sandberg’s fund, Sandberg Bernthal Venture Partners and existing backer Neo.
Flint’s goal is to create websites that continually optimize themselves, conduct their own A/B testing, and dynamically learn from each visitors and market trends, comparable to sudden interest in a specific keyword. Flint also goals to generate pages tailored to each visitor, much like Amazon displays personalized product recommendations.
Techcrunch event
San Francisco
|
October 27-29, 2025
That said, Flint’s technology is not ready to do all of it yet. “For now, users still have to tell us what they want to build,” Lim said.
In its current form, once parameters are set, Flint can mechanically generate web site design and layout, interactive elements (comparable to tables and buttons), in addition to offer form tracking and ad optimization. Lim says the platform can do all this in “about a day,” though she didn’t provide more details.
“At this point, customers are providing their own copy,” Lim said. She added that while Flint’s content writing feature will arrive in about a 12 months, future versions will provide customers with the option to write text via artificial intelligence.
Even without this, Lim says having a website ready with all the obligatory components in one day already saves her clients a lot of time.
Flint says he doesn’t design websites or “vibe code” anything. For existing websites, the technology analyzes the look and feel to build and deploy fully coded websites that adhere to the design.
The startup already cooperates with clients comparable to Cognition, Modal and Graphite, for which it has created working websites. You can see Windsurfing is here, Modal websiteAND this is what graphite looks like.
Flint’s ambition is to help marketers at high-growth startups and Fortune 500 corporations increase the visibility of their websites and create content.
The focus on selling to CMOs is what made Lim so excited that Sheryl Sandberg joined as an investor. “I like to think of her as someone who has made an impact on the way we make money online over the last decade,” Lim said.
According to Lim, Sandberg immediately understood Flint’s vision. “I showed her this deck and told her how, in my personal experience, it took five teams three months to create one A/B test just to increase our Google Ad conversion by 10%,” Lim said. “And then she stopped me, [and] she said, “Michelle, it must have taken 140 people at Meta to do this.”