I understand it's about cognitive costs. What I propose makes the cognitive load of actual monetary payment a single decision per month. I pay X. Let the aggregator and the publishers automatically figure out every little fractional detail.
If the objective of paying is to make the web better and reward quality content, you don't want two pages of clickbait to pay more than one page of watergate-level investigative reporting.
And any automated system to tell clickbait from quality journalism will incentivise people to trick it.
The original argument was essentially that there was a minimum transaction size that was cognitively viable at scale.