The new open-source economics nobody is writing about

Maintainers are quietly re-licensing. Not because of VC pressure — because agents read their code at industrial scale and take all the oxygen.

The narrative for the last decade of open-source license changes was straightforward: a startup builds something useful, venture capital demands monetization, the license switches from AGPL to something more restrictive, Hacker News has a weekend of indignation. Redis, Elasticsearch, MongoDB, HashiCorp — same playbook every time.

The new pattern is different. Maintainers with no VC pressure — hobby projects, academic tools, research utilities — are re-licensing specifically to exclude AI training and agent-driven usage. The reason is simple: their docs and source code are being hit by agents hundreds of thousands of times per month, rephrased into chatbot answers, and the humans who used to show up in their GitHub Issues have stopped. The projects are being used more than ever. Nobody is reading the actual site.

This is, mechanically, what agent-native commerce is trying to solve. If an agent's visit paid the maintainer a fraction of a cent, the flow reverses: more agent traffic becomes a funding source instead of a burden. The license restriction becomes unnecessary. The maintainer doesn't have to pick between "share knowledge freely" and "eat." That's the pitch. The next five years will determine whether it actually works.