In a September 25, 2026 post, a user recounts a conversation with a large data-labeling business that predicts Fortune 1000 enterprises—not labs—will provide most of its revenue in a few years. The business also believes every company will want to own its intelligence, without necessarily using open-source models.

The account predicts that internal evaluations—tests of AI agent performance—will become companies’ main proprietary intellectual property, citing performance gains from properly setting up and running evaluation environments. It also argues that most enterprises haven’t moved beyond coding agents largely because they lack the evaluation infrastructure to make agents perform well outside engineering.