The way we browse the web is going to change. It is 2026, technology moves on every week, and people now turn to AI en masse to answer their questions. I hardly ever use Google myself anymore.
Yes, this is a turning point, a moment when the way we browse is being rewritten: it calls for boldness and genuinely avant-garde experimentation. LLMs may have transformed how we access information, but text-based dialogue is still an ergonomic dead end when it comes to navigating the web. We are, first and foremost, visual and sensory creatures. That is what explains why Instagram or TikTok have been such huge hits: no cognitive friction, no slowdown, whereas chatting with an AI requires constant attention. So how do we imagine tomorrow's interface? How do we make the web truly "magical"? These are the questions driving me to test Jev today.
Developed by TypeSafe AI, Jev is an artificial intelligence model. Jev does not generate text. It is a "System 1" model used to make structured decisions. When you land on the page, the website asks its questions to Jev: it evaluates the situation in a few hundred milliseconds and returns a probability for each option. No thinking time like with Claude or Gemini, no text to process: Jev delivers its decision in a fraction of a second, which can be translated into action within the user journey.
Let’s put ourselves in the situation. You are a regular customer of a clothing brand; yesterday you browsed "by hand", clicked here and there, discovered a new collection, clicked again… Today, as long as you are logged in, the site knows you, knows your size, your measurements, knows your tastes and your past purchases. The page greets you by your first name. Jev has ranked the new collection in a fraction of a second, with a probability score, based on what is most likely to appeal to you. Another example still: you have planned to buy a few pieces for your summer wardrobe, for that holiday in Tuscany. The site knows it, because you connected your calendar to your customer account, and Jev turns this information into recommendations. At the bottom of the page, summer outfits are immediately available, based on your tastes.
Film transcript
Demo film, 1 min 33 s, English voice-over: the Lena Paris interface composes itself for Carmen. Timestamps are in minutes and seconds.
-
On screen: Black screen with the “graphen studio” wordmark; a white curtain opens onto the Lena Paris home page.
-
On screen: Lena Paris home page: the welcome messages are written word by word in time with the voice, then the grid of new arrivals (four coats) is revealed.
Voice-over: “Welcome back, Carmen. Here are pieces from the new collection we think you’ll love.”
-
On screen: The page invites Carmen to connect her calendar: the text is written on the left, a portrait of a woman in sunglasses fills the right.
Voice-over: “Share your dates. Connect your calendar, and let us prepare for the key moments of the year, curating the ideal wardrobe in advance.”
-
On screen: Trip to Tuscany: the message announces a summer selection, then four summer outfits appear.
Voice-over: “Your trip to Tuscany starts next week. Here’s what we’d pack for you.”
-
On screen: The linen shirt and palazzo co-ord is chosen; the other outfits fade away, then the page shrinks onto a grey background before opening the product page.
Voice-over: “Linen shirt and palazzo co-ord, excellent.”
-
On screen: Product page: photos of the outfit (white linen shirt, grey linen palazzo trousers tied with a silk scarf, close-up of the fabric), a price of €2,450 and an “Add to bag” button. A text-free myLENA panel pulses in time with the assistant’s voice as it describes the trousers.
Voice-over: “Crafted in Italy from pure linen, these palazzo trousers feature a wide silhouette and an airy, effortless drape. The high waist is accentuated by subtle front pleats, while the concealed fly and slant pockets preserve beautifully clean lines.”
-
On screen: “Complete the look”: three suggestions appear as the voice names them: a linen shirt, a silk twill scarf (€420), then a silk voile stole labelled “For your wardrobe”.
Voice-over: “To complete the outfit, you can explore the linen shirt in detail, or the scarf matching the set. For your wardrobe, I’d suggest another option. This silk voile stole. It’s an effortless match. It works naturally with this look and pairs just as easily with your cotton vest purchased last month.”
-
On screen: Fade to white: TypeSafe AI logotype, then the “graphen studio” signature.
The idea is to go far beyond classic product recommendation, which is already relatively widespread and automated across most of the web. What we are looking to explore with Jev is the concept of Generative UI: a "magical" interface that composes itself in real time according to each user's profile. The storefront becomes a starting point, built from what the brand knows about you and your inspirations. Of course, this ultra-personalization of the experience is not just about fashion. It applies to media too: newsrooms could write articles in the tone and from the narrative angles each reader cares about, based on their comments and likes, in short, on their personality. We can also imagine online stores connected to your Pinterest account, offering you the styles that match you best. Jev is capable of quantifying all of this and bringing up the most personalized result.
In this quest for efficiency, LLMs and Jev are not rivals. On the contrary, their combination is an incredible opportunity: an LLM that thinks and dialogues with the customer, and Jev that decides and composes the interface live. Provided, in my view, that we move away from chatbot interfaces, which I find completely outdated, to integrate conversation right at the heart of the page.
Even so, let’s remain clear-headed: we are in the middle of experimentation. Jev itself is only in early access, and what I am describing here is a matter of testing and sampling. We can test these living interfaces on certain pages, with a portion of visitors, adjusting as we go. It is not about replacing the entire website. A more traditional interface should stay in place, with its menus, search bar and categories, for those who like to browse and discover things for themselves. Personalization is a starting point, not a destination: everyone should be free to leave it behind at any moment and explore their own way.
Perhaps the web of tomorrow will sit somewhere between conversation and navigation: an interface that anticipates our needs without ever boxing us in.
Graphen Studio · Interaction Design