The web was built for humans. After thirty years of largely uninterrupted service, that assumption is being quietly disrupted by a new user base that doesn't click on links, doesn't watch ads, and doesn't sleep. AI agents have stopped writing copy and started doing things: navigating websites, filing forms, processing transactions, and sustaining hours of work unsupervised. The plumbing of the human web was never built with bots as customers, and the infrastructure pivot is now underway.

Two data points from the past fortnight. Alibaba's Qwen3.7-Max, released on 20 May, ran unsupervised for 35 hours and rewrote a piece of software to run ten times faster on hardware it had never seen before. This isn’t a chatbot, it’s a freelance engineer. Meanwhile Parallel Web Systems, founded by ex-Twitter CEO and Stanford CS PhD Parag Agrawal, has just raised at a $2bn valuation building agent-native web infrastructure on the thesis that AIs will use the web far more than humans ever have.

🧐 What's in it for me? Less time clicking, more time delegating. The deeper consequence is economic. Every business model on the open web (ads, subscriptions, paywalls etc.) was built around capturing human attention. Agents don't browse for fun and don't watch ads, so the next decade's web economics will be rebuilt around access tolls, machine payments, and pay-per-fetch deals. Publishers and platforms haven't quite worked out where they sit in that yet.

💵 Out of the Lab: Almost every layer of the internet stack now has a startup trying to redo it for agents instead of humans. Search, browsers, payments, identity, content licensing: all up for grabs.

Until next time.

Like what you're reading? Share toast with a friend.