4 ms·
I think there is a socioeconomic problem regarding the wholly text-based environment: It's too good for the user so it can't exist! If you have it there are n
by safety1st 2y ago
I think there is a socioeconomic problem regarding the wholly text-based environment:
It's too good for the user so it can't exist!
If you have it there are no problems left for anyone to profit upon.
Something like this is nirvana for me:
* The email arrives in the client (which is my text editor/IDE/whatever)
* There is a macro or hotkey which transforms it into a piece of text that is a representation of a JIRA ticket (yes it's tragic that I have to use JIRA)
* I review, hit save and the JIRA ticket is updated/created
* There is another macro which transforms it into a Git commit, Slack message, etc.
But the world is not like this, hours of our work day are wasted manually copy/pasting, massaging and reformatting our information between different walled gardens and different user interfaces, imposing new cognitive load every time.
Let me just have a set of templates in my text editor which do this transformation, I review it and hit save.
Emacs is the only application which has any meaningful traction and approaches this type of scenario: a generalized collection of buffers where your computer is just pushing text around for you and interacting with other bits of software to persist those changes in different ways (generally easier if text is the interface for those applications too). I think it's no coincidence that Emacs comes from GNU.
If all these things are just text in my main environment, there aren't many problems left that people can sell me. Import/export of proprietary file formats is a classic way to lock people into an ecosystem. SaaS APIs are a more modern one. If all this is just text files editable by me, if the interface is universal, the lock-in fails.
Give me text I can transform in my powerful editor. For a power user at least, everything else just gets in the way.
This is a place where I think LLMs might develop some legs by the way. Most applications need to accept text or at least keystrokes from the user sooner or later. LLMs seem to be pretty great at transforming text between two formats/vocabularies/whatever - maybe one working for me can do an end-run around the walls that proprietary systems erect, and make it easy to perform these transformations I'm talking about. Maybe your operating system vendor sees this as being the next step in the arms race between the different companies competing for your attention.