His talk (you can see the slides at the bottom of this post) was titled “Why Widget Is the Wrong Word and Why it Matters,” but it was really about why widgets aren't working today. ...as widgets started to be used to display other web content...they became 'relegated to the sidebar and increasingly seen as ad units and increasingly ignored.' ...But they shouldn’t be ignored, Fred argued. They should be integrated into the flow and experience of the page. Developers, he said, need to put more focus on widget user experience.
I believe Fred is more or less on point with his observations, but Nate's point about "context" is even more essential. In the long term widgets only make economic sense within the context of editorial experiences on websites outside the context of traditional social networking experiences. In fact the basic building blocks of content-oriented websites will increasingly be "widgets" (e.g. blocks of modular, interactive content informed by social graph data).
As Fred points out in his blog it's all about mashing up web services, and so the components of the widget, itself, must be modular and flow seamlessly with the page. Therefore a new generation of "widget builders" is necessary for the task, which is something we've been working hard on at KickApps. Moving forward, a single widget may draw on data from multiple web services and often won't resemble the boxy things we now call "widgets".
But it's not just the definition of "widgets" that needs an overhaul to be relevant outside of the context of traditional social networking experiences. The "social graph", itself, needs break free from traditional social networking websites, and the reasons go well beyond "profile portability" (which industry people often make way too much of).
Widgets (and all content) increasingly need to be contextually-informed to be relevant, and that transformation of content can only happen with access to contextually relevant social graph data. There's no doubt that social graph data will be used to inform (socialize, personalize) editorial experiences on sites across the entire web---miles from the nearest social networking portal. The problem for publishers is that existing uber-social graphs are simply not up to the job of serving individual websites. Friend Connect and similar efforts may prove to be valuable for certain limited tasks as a kind of mega-phone book (e.g. portability), but the future will belong to web publishers that learn how to organize and make use of their own audience data, collected within the context of their own user experiences.
End Note: KickApps provides the industry's first Social Graph Engine (tm) for Publishers.