Great question, and one that certainly doesn't have a straightforward or trivial answer. It's definitely more of a social challenge than a technical one - making publishing free/open won't do anything to fix incentives on its own.

My hunch is that change to this system will come from the outside. It's too risky of a career decision for a tenure-track professor to start publishing on PubPub (or any open/new system). But, there are lots of people who aren't playing that game. Lots of people who are doing science outside of academia, at a corporate R&D position, or for the sake of education, etc.

The most important step is to show that open publishing works. If we can work with these early adopters and show that conversations are more rich, or results more reproducible, we can start to go to universities and grant agencies and advocate for them to require open publishing. The first day that a university hires a professor or an agency rewards a grant based on the history of openly published work, will be a turning point. I hope it will be similar to the first time a software dev was hired for their Github profile, rather than their CS degree.

Today, software companies hire on experience. A university degree can show that, but so can major contributions to an open-source project. I hope science can become the same. Whether you're a PhD out of a great program, or an high-school drop out that has committed her life to rigorous experimentation, your demonstrated experience should be what you're hired on, not the list of journals that have found it in their interest (many of them are for-profit) to include your work.