UPDATE: I posted some further notes on the fact that Twine is in beta, and what "beta" actually means and why we are in beta here.
Marshall Kirkpatrick wrote a critical review of Twine today that identified several known issues the team is working on. These are points well-taken -- we certainly understand that Twine is still a work in progress and there are many areas where we can improve usability. After all, Twine is still in private invite-only beta and is not a finished application yet. There is much that is still under development and we are learning from our users everyday.
However, we have also been getting quite a lot of very positive feedback from our beta testers as well. Twine is already quite useful and works surprisingly well on a wide selection of Web content today, as our growing beta user base can attest to.
So on balance, while Marshall points out several issues we are aware of and are working on, there is much we are proud of in what we have been able to accomplish so far.
But I want to address some of the specific points Marshall made. Marshall pointed out the following issues:
- Sometimes Twine is unable to auto-summarize the content of some pages it sees, when they are added to the service.
That's true -- it's sometimes hard for Twine to identify the "content part" of a page when the page has complex structure (including tables, Flash, Ajax, frames, multiple DIV areas, etc.). In the meantime, Twine does actually do a good job on things like Typepad Blogs, the Wikipedia, Youtube, Flickr, Amazon books, Wordpress, and most sites that have relatively standard page structure and/or metadata. That said however, we are working on making Twine smarter so that it can do a better job, even when there is uncertainty about the content and structure of a page. As Marshall points out this is a hard problem because there is so much non-standard content on the Web, but it's not an insurmountable one. Twine will steadily improve over time on this front.
- Twine doesn't recognize article authors as related people.
Actually, if Twine can see the author's name, it will recognize them as a related person. But the author's name is not always visible on the article. It would be easier to manage this if there was better metadata on pages, but until that happens, the natural language approach is the main option, and it is not always perfect.
- Marshall mentions that he thinks Twine could be better organized.
Marshall mentions that he had a hard time getting oriented and finding his way through the application because there is so much there. One of the challenges we have is simply educating users about how to Twine and what it is capable of. In addition there are many improvements we know we can make to the user-interface and information design to make it easier to figure out.
Marshall also asked for RSS feeds and visualizations.
RSS output is already supported to a limited extent and we will have more support for it next month. We are also planning to add RSS input as well in coming months.
Regarding visualizations, we've done a lot of work on visualizations in the past. Our feeling is that they usually don't add much value, other than being eye-candy. However, we will be opening up our API's eventually to allow others to make all the visualizations they want. If someone makes a really useful one, perhaps we'll include it back into Twine.
Finally, I would also like to correct one thing that Marshall mentioned: We are not in fact going into general release next month -- we are just starting to let more people in from our waiting list to continue to help beta test Twine. There will still be a members-only policy in effect for several more months. The full public opening (when Twine will be opened to non-member guests, and search engines, etc.) will be in the summer timeframe. Even then, Twine will still be in beta. There is a good year of additional work to do on Twine before it will be fully "baked," to use Marshall's term. Between now and that time we will be working to improve (and finish) the app, in partnership with our beta community.
In closing, as I have said many times, Twine is still an early Beta and we have to keep expectations in line with reality. Twine is already far and beyond what any other semantic app I know of is capable of, but that still isn't good enough. We have to push further and focus more on usability. We are opening it up early in order to get feedback and more help testing and guiding the direction of the app from users.
Hopefully as we work on Twine further, and we move out of Beta, Twine will eventually meet Marshall's high expectations. Meanwhile, his comments are helpful in that they do give us feedback about what aspects of Twine we need to focus on more as we head towards a more consumer-friendly application.