1. Coframe Series A brings in $25 million
Coframe, a San Francisco artificial intelligence startup, announced a $25 million Series A round on November 9, 2026, co-led by Khosla Ventures and Hanabi Capital. The Coframe Series A lifts the company's total funding to $35 million since it was founded in 2023, and it will finance software that generates and adjusts web pages for each visitor in real time.
The new money arrives roughly two years after the company's previous financing, a seed round of about $9.3 million that closed in October 2024. Taken together, the two rounds account for nearly all of the $35 million the company has raised in its three years of operation. Josh Payne, who founded the business, serves as its chief executive.
A Series A is typically the first large institutional round a young company raises after it has an early product and some paying customers. It is the stage at which investors stop funding an idea and start funding a business they expect to scale. For Coframe, the round signals that two venture firms were willing to commit meaningful capital to the proposition that websites should no longer be fixed objects that a team edits by hand.
2. What does Coframe do?
Coframe builds software that creates, tests and refines websites, landing pages and advertisements without waiting for a person to direct each change. The company's pitch is that every visitor can be shown a page shaped for them, and that the page keeps improving for as long as it is live.
Under the hood, the platform relies on specialized transformer models, the same broad family of neural networks that powers modern language and image systems. Rather than using a single model to do everything, Coframe runs several AI agents that cooperate on a task. The agents split responsibilities in a way that mirrors a marketing department: one group of functions covers strategy, another covers design, and another covers data science.
That division of labor matters because improving a web page is not a single skill. Someone has to decide what to try, someone has to produce the new layout or wording, and someone has to read the results and judge whether the change actually helped. In a conventional company, those jobs belong to different people who hand work to one another over days or weeks. Coframe's system assigns each of those jobs to software and lets the handoffs happen automatically.
The output is a loop that never formally ends. The agents build a version of a page, measure how visitors respond, adjust it, and begin again. The company presents this as an ongoing process that runs in the background rather than a series of separate marketing campaigns with start and end dates.
3. How the model departs from A/B testing
Traditional website optimization has followed the same routine for many years, and Coframe defines itself largely in opposition to it. In a standard A/B test, a marketing team prepares two versions of a page, sends a share of visitors to each one, and waits until enough data has accumulated to say which version performed better. The winner becomes the new default, and the team starts planning the next test.
The method is reliable, but it is slow and narrow. Each experiment compares only a small number of options, each one requires people to design and approve the alternatives, and each one needs time before the numbers are trustworthy. A company that runs tests diligently might still evaluate only a limited set of ideas over the course of a year, and every visitor in the meantime sees one of a handful of fixed pages.
Coframe's system is designed to remove the waiting and the manual steps. Because the agents generate the variations themselves and evaluate them continuously, the cycle of proposing, testing and choosing is no longer tied to a team's calendar. The company describes this as a break from the compare-wait-pick-repeat rhythm rather than a faster version of it.
The second difference is personalization. An A/B test looks for the single best page for the average visitor. Coframe's stated aim is a page tailored to the individual, which means the system is not searching for one winner at all. It is trying to match content to the person in front of it and to keep refining that match as more people arrive.
4. The idea of a living internet
Behind the product sits a broader argument about how the web has been built. Coframe's founding premise is that websites have stayed static for too long. Most are designed once, revised occasionally and then left largely alone, even as the audience, the products and the market around them change. The company compares that pattern to aging hardware, suggesting that a site left untouched dates as quickly as an old flip phone.
Coframe calls the alternative a living internet: a web in which pages keep evolving in response to how people use them. In this picture, a website is less like a printed brochure and more like a service that adapts with every interaction. The company's product is meant to be a first practical step toward that goal, applied to the commercial pages where small improvements translate directly into revenue.
The framing is ambitious, and it reaches beyond what a marketing tool usually claims. It also explains why the company talks about websites, landing pages and ads together. All three are surfaces where a business presents itself to a potential customer, and all three have historically been produced by hand and updated on a schedule. Coframe treats them as one problem that the same autonomous loop can address.
5. The results Coframe reports, and their limits
Coframe supports its case with performance figures drawn from its own customers. The company says its platform has generated more than $300 million in incremental revenue for the businesses that use it. Incremental revenue refers to sales a customer would not have made without the tool, so the figure describes gains for Coframe's clients rather than Coframe's own income.
The company also points to results at individual customers. It cites a 158% improvement in conversion at Dropbox, the file storage and collaboration company. A conversion is the moment a visitor takes the action a page is built to encourage, such as signing up or making a purchase, and the conversion rate is the share of visitors who do so. A lift of that size would mean a page turned visitors into customers at more than double its earlier rate.
These numbers deserve the caution that applies to any vendor's account of its own product. They are reported by Coframe and have not been independently verified. Outsiders cannot see how the baseline was measured, how long each test ran, or how much of a given improvement would have happened anyway. Headline results from one customer also say little about what a typical customer should expect.
None of that means the figures are wrong. It means they are marketing claims until a third party confirms them, and prospective customers and investors will weigh them accordingly. The size of the Series A suggests the lead investors found the evidence persuasive enough to back.
6. Why the round matters for marketing teams
Funding of this kind points to where investors believe marketing work is heading. Optimization has long been a labor-intensive discipline, staffed by designers, copywriters, analysts and engineers who coordinate each experiment. A system that performs those roles with cooperating agents changes what a company needs people to do. The human contribution shifts from producing each variation to setting goals and reviewing what the software produces.
It also changes the economics of experimentation. When each test requires staff time, companies ration their tests and focus on the pages with the most traffic. When tests are generated and assessed automatically, the cost of trying an additional idea falls, and more of a site can be put under continuous review. For businesses that depend on online sign-ups or sales, even modest gains in conversion compound across every visitor.
The approach raises practical questions as well. A page that changes for each visitor is harder to audit than one fixed design, and brands generally want consistency in how they present themselves. Companies adopting autonomous tools will need to decide how much freedom the software has and where human approval is required. Coframe's description of its product centers on autonomy, and how customers balance that against control will shape how widely the model spreads.
7. What comes next for Coframe
With $35 million raised in total, Coframe now has to show that its early results hold as it takes on more customers. The company's claims rest on a small set of named examples and one aggregate revenue figure, all self-reported. Broader adoption would provide a larger and more varied record against which those claims can be judged.
The competitive setting is also likely to tighten. The idea that AI can write, design and test marketing material is not unique to one startup, and established optimization tools have their own reasons to add similar capabilities. Coframe's bet is that a system built from the start around autonomous agents will outperform products that add automation to an older, manual workflow.
For now, the Series A gives the company the resources to pursue its stated goal of a web that updates itself. Whether the living internet becomes a common way of building sites or remains one startup's description of its product will depend on whether independent evidence eventually matches the numbers the company has put forward.