The latest changes at Medium apparently involve the shut-down of some of the site’s magazine-style micro-sites or publications, such as The Archipelago and Re:form (the latter of which was a joint venture with BMW that depended in part on sponsored content). Other “collections” have been merged, according to various reports, including Matter, which is effectively no longer considered its own standalone site, according to a recent report.

All of this is part of what Williams has described as a shift in emphasis, away from just traditional-style publishing, and towards becoming more of a Twitter-style social network. In a post he wrote about the changes, the Medium CEO said the site “is not a publishing tool. It’s a network. A network of ideas that build off each other.” But despite these protests that everything is going according to plan, there seems to be a fair amount of angst both inside and outside the company.

To be fair, Williams hasn’t done himself any favors when it comes to conveying what Medium is or wants to be. He doesn’t give a lot of interviews, and when he does talk about it, the purpose of the site seems to shift and change. In the early days, he said that Medium wanted to develop a kind of stable of digital magazines — and that’s very much what it appeared to be doing, as it acquired the Kickstarter-funded science magazine Matter, and more recently brought veteran tech writer Steven Levy in to start a site called Backchannel.

Then Medium said it wanted to be part publication and part platform: it would start magazines or allow others to start them, and in some cases hire writers to produce content, but then it would also be an open platform where anyone could post their thoughts, and editors would collect content of both kinds. Writers like Clive Thompson and Paul Ford were invited to become part of “collections” like The Message (which has also been restructured, reports say) and others started up devoted to specific topics, like the female-focused Ladybits.

As Medium experimented with paying writers and editors based on the traffic their pieces generated, however, a tension arose between those who saw the site as a place to write at length without having to worry about such concerns, and the pressures of being a startup that had raised $25 million from venture-capital investors. Ladybits editor Arikia Millikan quit running the site because of some of these pressures, and wrote about her decision (on Medium, of course) and Medium later moved away from paying staff based on traffic.

Now Medium wants to stress the social aspect of writing — which, to be fair, was always part of how Williams talked about the site (as he noted in a response to the BuzzFeed post on the changes). The preceding round of changes at Medium involved making it easier for people to post short items, Twitter-style, and to turn these short items into responses to other people’s posts. In effect, the site implemented comments, but didn’t call them that. Each comment is a post of its own, a little like the way that Gawker Media gives commenters their own blog using its Kinja platform.

For a site that wants to reinvent media in some way, this kind of experimentation isn’t just interesting, it’s crucial. And I’ve written before about how new entities like Vox need to be given more leeway to experiment and fail and try new things — rather than being expected to somehow save the existing media industry with all its flaws and antiquated processes.

Is it frustrating sometimes to not know what Medium wants to be, or how it plans to get there? Sure it is (and I have also written posts about how Medium needs to figure out what it wants to be when it grows up). But if Williams has decided that he’s more interested in experimenting with interactivity and less in becoming a stable of magazine-style sites publishing fairly traditional magazine-style content, then I can’t say I blame him.

The site’s new approach may disappoint writers and editors who were hoping that Medium would somehow figure out how to rescue the traditional publishing industry, but so be it. Ultimately we are all going to learn more by watching the site experiment than we are if it just replicates an existing business in digital form. We’ve got quite enough of that already.

Subscribe to Data Sheet, Fortune’s daily newsletter on the business of technology.