I'm not planning on ending support of the local-brain based Hal. The next version of Hal will let the user have the option of using the internet-based brain or the local one. So don't worry about losing the brain you have now, you'll just be getting a new option.
Right now I'm experimenting with having Hal's new algorithms analyze the conversations going on Twitter. There are over a million conversations going on Twitter every month and through the Twitter API I can analyze many of them. After filtering the conversations for quality, it seems like it will be a huge source of knowledge for Hal. Hal will be able to talk about current events going around in the world and, pop culture, and since Twitter conversations are tagged by region Hal may be able to focus his knowledge to a region near you.
I'm looking into ways of allowing both client-side and server-side scripting so that many of the cool plug-ins developed can continue to work seamlessly. I am planning on making a web browser based version of the Hal Brain Editor and I'm hoping of having an upgrade path which will allow uploading your local brain into the "cloud" if you so choose.
There is still much work ahead to make this a reality but I am striving to make it as smooth a transition as possible.