What AI Shopping Assistants Miss: The Few Seconds Before You Decide to Buy

AI shopping agents share one assumption: you already know what you want to buy. Enter 'running shoes under $120,' and it executes. But a lot of shopping does not start that way. @0xMovez writes in a tweet that shopping often begins with scrolling, seeing something, and thinking 'I need this.' Agents execute intent, but no one is addressing the moment before intent appears.

Dalet Labs just launched Scene, which it calls Generative Commerce. Founder Omer Bar Tal puts it directly: it is like walking into a mall built around your taste, and everything can be bought inside the app. There is no prompt and no chat box in the product. The feed is made of realistic scenes generated around your taste; every product in the image is actually tappable, and items across brands and categories can go into the same shopping bag and be checked out in one go. What you linger on and tap influences what gets generated next.

The product shown on the website is a scene image plus tappable products, for example:

This is rather different from the agentic commerce common today. The latter is good at replenishment: paper towels, laundry detergent, the cat food you bought last time. With a clear goal and clear parameters, the agent compares prices and places the order. Discovery shopping is the opposite: users often see something first, then develop intent. Asking an agent to execute a command that does not yet exist is inherently awkward. @0xMovez goes further: apart from restocking paper towels, Generative Commerce will eat Agentic Commerce. That judgment may not be right, but it points to the problem.

Dalet Labs has raised a $7 million pre-seed led by Lightspeed, with angels including SV Angel, Hersh Desai, @FEhrsam, @rauchg, and others. The team includes research and engineering people from Google, Runway, and MIT. Early testing has been running for several weeks; it is currently invite-only on iOS, and you can apply for early access on the website.

Here are a few more scenes and products from the website; the items in the images all correspond to real products and prices:

The risks are also obvious: products in generated scenes must match real inventory, prices, and sizes; recommendations can easily become narrower and narrower; and if fulfillment and returns for cross-brand checkout break down, no matter how good the content looks, the experience will fail. But the product has a sharp insight. That moment before the shopping cart may be worth more than the search box.

Source: @0xMovez tweets, Dalet Labs website.

https://x.com/0xMovez/status/2115484262732317044

https://daletlabs.com