Where will next-gen open model data come from? Task commons doesn't exist yet

NeelRajani_ · x · 2026-10-16

@dlwh argues: the internet offered a remarkable commons for pretraining, but posttraining consumes tasks, not documents — and that commons doesn't exist yet.
Neel Rajani pushes back: pretraining data was far from a true commons; huge IP theft was involved, many writers were never compensated and suffered real earnings losses.
The debate centers on where data for the next generation of open models will come from and how to frame its legitimacy.