Projects like this are so so close but fall just short of the ideal. I've been thinking about this for years but I have not been at the point in my life where I could implement my ideas which are these:
1.) A UI which is just a line/text field to enter commands. Something like the command prompt but which fuzzy matches commands like the mini-buffer in emacs or the omni text field in Chrome or Firefox or even Enso from a few years back.
2.) Each command is name spaced to an "agent" to avoid command collisions. For example agent 'jarvis' would have a set of commands it response to like jarvis/foo, jarvis/bar or jarvis/baz.
3.) The output of each command is a list of 0..N items/objects rendered in a master/detail view where navigation over the list shows a detailed view of each object/item in the list.
4.) An item/object can be anything from an email, rss entry, web page, graphic, tweet, contents from a text file. Basically anything that is renderable.
5.) The output of any command can be piped to any other command which is able to parse the list of items/objects from the prior command and render its own new list.
This UI paradigm seems to cover an incredibly large set of use cases. The only use cases I can think of which are not covered are those where the keyboard input device is not sufficient; such things as graphics manipulation where a mouse or pen & tablet are needed.
The frustrating thing for me has been to witness the vast number of systems over the years that have nibbled at the edges of this paradigm but have not gone all the way. What I'm talking about mostly here are the numerous launcher systems like Enso or Quick Silver or dMenu. All these systems have UIs very similar to what I'm talking about but they're restricted to launching existing apps and controlling the options exposed in menus of existing apps.
The other class of applications I've seen that come close are the ones like that mentioned in this topic. Applications like notty where the effort is spent trying to shoehorn extra rendering capabilities into a terminal emulator.
What I want is essentially a Grand Unified User Interface (GUUI) such that applications as we know them are done away with and we only deal with commands and output.
A system where I can type web/news.ycombinator.com and a one item list comes back with that first item selected by default in the details view. And that item is the front page of Hacker News. Then I could next type email/inbox and a list of emails in my inbox are rendered. And of course while viewing one of the items in my email/inbox I could type email/reply which would render a text area to reply to my previously selected email.
As I said earlier, the use cases seem endless and this paradigm seems like it would be incredibly efficient for those who can type well.
1.) A UI which is just a line/text field to enter commands. Something like the command prompt but which fuzzy matches commands like the mini-buffer in emacs or the omni text field in Chrome or Firefox or even Enso from a few years back.
2.) Each command is name spaced to an "agent" to avoid command collisions. For example agent 'jarvis' would have a set of commands it response to like jarvis/foo, jarvis/bar or jarvis/baz.
3.) The output of each command is a list of 0..N items/objects rendered in a master/detail view where navigation over the list shows a detailed view of each object/item in the list.
4.) An item/object can be anything from an email, rss entry, web page, graphic, tweet, contents from a text file. Basically anything that is renderable.
5.) The output of any command can be piped to any other command which is able to parse the list of items/objects from the prior command and render its own new list.
This UI paradigm seems to cover an incredibly large set of use cases. The only use cases I can think of which are not covered are those where the keyboard input device is not sufficient; such things as graphics manipulation where a mouse or pen & tablet are needed.
The frustrating thing for me has been to witness the vast number of systems over the years that have nibbled at the edges of this paradigm but have not gone all the way. What I'm talking about mostly here are the numerous launcher systems like Enso or Quick Silver or dMenu. All these systems have UIs very similar to what I'm talking about but they're restricted to launching existing apps and controlling the options exposed in menus of existing apps.
The other class of applications I've seen that come close are the ones like that mentioned in this topic. Applications like notty where the effort is spent trying to shoehorn extra rendering capabilities into a terminal emulator.
What I want is essentially a Grand Unified User Interface (GUUI) such that applications as we know them are done away with and we only deal with commands and output.
A system where I can type web/news.ycombinator.com and a one item list comes back with that first item selected by default in the details view. And that item is the front page of Hacker News. Then I could next type email/inbox and a list of emails in my inbox are rendered. And of course while viewing one of the items in my email/inbox I could type email/reply which would render a text area to reply to my previously selected email.
As I said earlier, the use cases seem endless and this paradigm seems like it would be incredibly efficient for those who can type well.