AI Ends The Internet*
*as we know it.
This Year The Web Turned 35! Happy Birthday, Internet.
Now imagine its skin changing: from pages and tabs to a single conversational + visual surface where agents do most of the work and we provide oversight. OpenAI’s recent moves point straight there. Pulse runs research for you while you sleep and hands you morning cards drawn from your context and connected apps — it’s the assistant as a proactive start screen, not a passive tool. In parallel, commerce is walking into the chat. Instant Checkout and an Agentic Commerce Protocol let people, agents, and businesses transact inside the conversation. And then there’s media. Multiple outlets report OpenAI is preparing a Sora 2–powered social app with a vertical “For You” feed, where the videos are AI-generated and the creation UX is as fast as the scroll. If that ships, your content supply becomes near-infinite and near-personal by default.
“Let’s be clear — what OpenAI is doing right now is ground-breaking. But not in the way most think. We’re clever animals who still run on Stone-Age heuristics; we save brain power, prefer shortcuts, and default to linear paths. So we’re slow to imagine interfaces that don’t look like what we’ve used for 25+ years and therefor we think very linear when it comes to transformational shifts, like the ones OpenAI is pushing for.”
Dominik Heinrich, Co-Founder Creative AI Academy
Forrester’s Emily Pfeiffer warned in an WSJ interview as agentic shopping lands that the merchant gets the sale … but they lose the loyalty. Brands are excited — and nervous — because the interface shift moves even faster than their operating model.
From Digital Goods to Consumer Intelligence Goods (#CIG)
We’re moving from static product pages to Consumer Intelligence Goods — offerings that ship with a live reasoning layer. A #CIG isn’t just content or SKU data; it’s a model-readable, agent-negotiable object that can be found, explained, configured, and purchased without a traditional website flow. Your “PDP” becomes an Agent-Ready Pack: claims, constraints, service/warranty rules, sustainability proofs, inventory, returns logic — all signed, versioned, and safe for assistants to act on. (This is exactly what Instant Checkout + the open protocol are pushing toward.)
The UI After the Internet As We Know It
I think 90% of interactions will be executed by ambient agents — an invisible layer around us — while our interface becomes a light verbal/visual confirmation surface: “Yes, do that; tweak this; explain why.” Wearables will be the bridge, but today’s “apps-on-a-display” metaphors are the wrong pattern. Even Meta’s new Ray-Ban Display glasses (a legit step forward) still mirror old UI conventions; what’s next is dialog-first overlays that show only the rationale, options, and risk at the moment of choice.
What changes in consumer behavior
From search to brief. People will brief assistants (“plan my week, optimize my meds, find me a jacket I’ll actually wear”), not navigate menus. Pulse is already training this habit.
From ad impression to agent negotiation. Offers need to be machine-understandable (eligibility, caps, evidence), so agents can assemble carts and justify the choice. That’s the point of an open commerce spec.
From loyalty to utility. Loyalty shifts to the outcome experience. If the agent consistently nails fit, comfort, price, and support, the brand gets the sale — but must re-earn relationship in post-purchase service.
The Hard Truth For The Internet
The homepage won’t disappear tomorrow, but its center of gravity will. By the time Sora-style feeds normalize and agentic checkout is multi-cart, a large share of discovery will start inside assistants — not on the open web. The brands that win will create real-life experience that are grounded in contextual relevance and more importantly treat products as intelligent objects with service baked-in that goes beyond FAQs.
This shift is exciting, but requires designing for the unknown. It will absolutely create better experiences when done right. It will also punish shallow claims, fuzzy data, and generic design. Let's design for human connection not for purchase decisions.