Components Will Kill Pages

The internet is changing. That is undeniable. For the first time in over 20 years websites are also about to drastically change. We are familiar with browsing the internet website by website, one page at a time. Lately with tools such as ChatGPT becoming available, our TikTok attention spans demand immediate results and ChatGPT delivers just that. You open a window, type in a text input anything you want to know and – boom, an answer appears catered specifically to you. Our remaining brain cells slam their glasses together to celebrate the quick dopamine hit, accessing information at minimal effort. Times are changing.

Frontend engineers have been solving problems at the user journey level for the longest time, optimizing how users navigate from page to page – or feature to feature. We’ve asked ourselves thought provoking and hotly contested questions such as “What should the signup funnel include” and “how do we get the user to the value add the quickest” many times in the past twenty years. Now I’m saying none of that will even matter. Will it even be a user signing up for your product, or AI on their behalf?

Snapping back to ChatGPT and how it’s shaping how users think about interfacing with the internet. In a world where we can type anything into a text box and get the information back instantly we are circumventing the need to visit websites altogether. Bots will continue to scrape your vast amount of pages that exist but users will continue to grow more lazy. Obviously it’s not advantageous to AI chat applications to link users out to websites, they want to keep them on their app just like X penalizes links in posts and mobile app developers seldomly link outside of their app. What if there was a way to still introduce visual components to users through chat interfaces that let them interact with the brand?

Let’s take Cloudflare for example. For all the Workflows I have deployed on the platform it would be kind of gnarly for AI to try to visualize for me how they may work (steps, retries, try/catch, etc.. trust me). If AI wants to own the internet then having a method for chat applications to understand components exposed from the website in question (e.g. Cloudflare) and know how to call for and render it inline to the chat response – that would be something else. No requirements to redirect the user out of the app and to that website, and still providing a best-in-class user experience. AI chat applications win, and so does the brand that gets to keep ownership of how it tells the story of its product to its users.

With libraries like json-render and component libraries like Kumo you can have UI components be represented in JSON such that AI knows how to create generative UI’s. An application can understand what components are available from the pool of components within Kumo and using json-render AI can understand which components support children, what props they have, and how to stitch them together to give unique responses. For example if I ask for one of my aforementioned workflows to be visualized to me and Kumo had a FlowDiagram component that knew how to construct it, ChatGPT could use this information to respond to the user with an interactive workflow viewer with Cloudflare’s brand.

Imagine your own product. You likely won’t expect users 5 years from today to navigate 5 pages deep, apply filters and sort data just to try to derive the answer themselves… will you? My hope is that you will embrace the change and have a first party experience where people can ask a question and get an answer. Users can trigger actions with a few words and get results in seconds. AI will be the new steering wheel of the internet, it won’t be navigating pages.

Now is the time to invest in high quality components. If you’re not using an AI first component library or thinking about how AI could consume or create user interfaces with your branding then now is the time. You’re not behind (yet) but in today’s world where the world changes every month, it’s best to be ahead.