Base44’s CEO says Base 1 fights AI website sameness

Base44, acquired by Wix last June for $80 million, has released its own large language model, Base 1. The San Francisco startup says the move is meant to stop AI-coded websites from looking interchangeable—offering users model choices like Claude’s Opus 4.8, F

A new app-building prompt is already changing the vibe of “vibe-coded” websites—starting with a decision that looks small on the screen but bigger behind the scenes.

Base44, a San Francisco startup, said Monday it trained and released its own large language model called Base 1. Before users enter a prompt to build an app, they can now select Base 1 from a menu of AI models, alongside options including Claude’s Opus 4.8, Fable 5, and OpenAI’s GPT-5.5.

The pitch is straightforward: stop the cookie-cutter look.

Shlomo. the company’s CEO. told Business Insider that part of the reason Base44 developed Base 1 was to combat the telltale generic appearance that can show up in many vibe-coded products. His concern is specific. Using frontier models to generate websites can leave people feeling like they’re getting the same user interface over and over. “Everybody feels like they’re getting the same UI when they’re coding with the general models,” he said.

Base44 is not starting from scratch. Last June, the company was acquired by Wix for $80 million, bringing it into Wix’s website-building arsenal. Shlomo said Wix has a large team of designers, which creates a lot of data for the model to train on.

That training process is where Base 1 is meant to diverge from the typical approach. Shlomo said the Base44 team will conduct “reinforcement learning” on the new model—prompting it to keep generating designs that look new and unique.

While Base 1 is “not yet there. ” Shlomo said his goal is for it to “create something that looks uniquely different” each time it generates a user interface. He said the team began working on Base 1 about six months ago. but breakthroughs in the last few weeks allowed the company to release the model sooner than planned.

For Base44, uniqueness isn’t just a design preference—it’s a competitive strategy.

Shlomo said generating distinctive designs will help set Base44 apart from competitors. Startups such as Lovable, Replit, and Cursor compete with Base44 in the same space.

The concern about sameness isn’t coming only from Base44. UI/UX experts and design gurus have warned about what they call the “AI slop” look in vibe-coded websites and apps. Paul Bakaus. the CEO of AI design startup Impeccable. said in a June interview with Andreessen Horowitz that common signs include beige or tinted backgrounds and sans-serif fonts. He also compared the resulting aesthetic to an “algorithmic Uniqlo or Ikea” look—design that isn’t necessarily bad. but not unique.

Shlomo framed the technical reason behind the problem in a way that connects back to the design criticism. Frontier models. he said. must be broadly good at everything. “from poetry to coding.” His view is that a general-purpose model can end up producing general results in each category. “And we think that if we take a model and we squeeze its ability to be really. really good at one use case. then we have a shot. ” he said.

The sequence is built into the product choices: users get Base 1 alongside other frontier options like Claude’s Opus 4.8, Fable 5, and OpenAI’s GPT-5.5, while the company argues that training for one use—creating interfaces that don’t repeat itself—is the key difference.

Base44’s message is that the web-building future doesn’t have to look copy-pasted. Whether Base 1 can consistently deliver on that promise—especially since Shlomo described it as “not yet there”—will be what users test next, prompt by prompt.

Base44 Wix Base 1 large language model vibe coding AI websites reinforcement learning user interface Claude Opus 4.8 Fable 5 GPT-5.5 Impeccable Paul Bakaus