Well, objectively speaking, it seems like the old model wasn't working. It wasn't scalable. Without standardizing something -- be it the buyer-description-input track, or the tasker-bidding-assignment track -- you have an ad hoc system that is not particularly efficient at allocating supply to demand. (Or even particularly good at standardizing supply and demand in the first place).

This new system seems to fit the classically snarky definition of a compromise: a situation nobody is completely happy with.

The article implies a third option: keeping the open-bidding system of old, but focusing on a vertical or two, instead of trying to be all things to all potential tasks. This might be a better solution, but again, you run into standardization and supply/demand allocation inefficiencies. Less of them, but they're still there.

There seems to be a good reason why "second-generation" sharing economy services, like Uber and Airbnb, do really well, while services like TaskRabbit don't. These newer services are on tightly controlled rails, they're standardized (or at least standardizing), and they're operating in specific verticals. They are in the variance-minimization business. Old-model TaskRabbit was in the variance-creation business.