The open internet may be giving way to software built for one person

Ben Thompson’s “end of the internet” is not a claim that networks disappear. It is a claim that the old bargain stops working. In the early web, anyone could operate a server and distribution was effectively free. Abundant publishing made Google’s discovery valuable; abundant social posting made algorithmic feeds valuable. In both cases, openness created opportunity and then competition at a scale that transferred the economics to an aggregator.

AI intensifies both failures. Economically, answer engines consume publishers’ work and attention without reliably returning the traffic that search once supplied. Technically, agents give attackers cheap, persistent iteration against software built on decades of human-authored assumptions. A hobbyist server that was merely inconvenient to secure becomes irrational to expose. Hosting, patching and defence consolidate into managed platforms.

Thompson’s more interesting argument is the countertrend. The same systems flooding public app stores with interchangeable software make it viable to build an application for one household, team or group of friends. The fixed development cost collapses far enough that no mass market is required. He compares this with the movement of blogging into paid niches and public social networks into private group chats: scale retreats, but smaller high-trust spaces become more useful.

This is a thesis, not an empirical forecast. Managed hosting already absorbed much of self-hosting before current AI, private software still inherits supply-chain and authentication risk, and millions of tiny apps could create a maintenance disaster. Central platforms may capture the tooling just as aggregators captured discovery. “Small” does not automatically mean sovereign.

Still, it offers a better frame than arguing whether AI creates or destroys software jobs. The unit of viable software can shrink. A workflow that would never justify a SaaS company may justify an afternoon of agent-assisted construction. The opportunity is not another universal app; it is a vast number of precise tools whose market is an association, a team—or one person.