i've been using a terraformed chromebook as my main machine for the past year. i don't have x11 running on it, just the framebuffer console, that's part of linux kernel. i have a desktop at the office and a pile of machines that i can pull from in a bind. i'm guessing this is rarely done anymore, since most packages that use framebuffer directly haven't been updated since about 2015. most activity (though not new development) is coming from raspberry pi users. i've written a dozen of variations on framebuffer writer at this point, in c, lisp and ada, so i can display visual data on screen. this gives me images, graphs and pdf documents. software that uses sdl 1 can theoretically run on framebuffer, but is rarely tested that way. gentoo packages for example usually hardcode sdl+x11 as a depedency. sometimes they work out of the box, but most of the time they require non-trivial modifications. i've gotten scummvm to work, dosbox is giving me issues, and i gave up on GrafX2. needless to say i can't browse present day web with lynx, which acts as a pair of "They Live" glasses. sites that you expect to know better, like github, reveal their true nature. i make a mental note of these sites, i think they ought to be shunned in general. there's plenty of web 1.0 resources to have to deal with this shit. there ought to be a solution to banking problem though. sometimes sites have apis, but i haven't explored that as an alternative. frankly writing json getters to a dodgy spec is not in the spirit of non serviam. i'd rather spend the time writing real code. i do work with this setup, but then most of my work involves writing up architectural decisions, reviewing other people's code, or doing refactoring. i've gotten into a habit of typing up a lot of changes over a course of several days without once running anything, and then testing the result on a work desktop. an approach that significantly reduces stress: i don't have to make it work while i'm thinking and i don't have to think when i'm getting it to work.