With all the recent trends around hardware tinkering, such as 3D-printing and cheap ARM-boards, I don't think such project would be entirely infeasible, or even overly expensive.
Maybe packaging it up as a tablet app would be a good first step in this direction?
I don't think porting to tablets (Android) would help that much. The reason for that is that Android is so vastly different from regular Linux that the work couldn't be readily reused. Also designing a good UI for touch-only would be very different from a keyboard-based UI. That is especially important if the hardware keyboard is one of the main selling points.
The development of such device could feasibly be divided to three parts, which could executed partially in parallel.
First would be the software, which I think would probably require the most work. I'd (re)write the whole UI, designing it from scratch to suit small screen. Most of the software work could be done on PC, which makes life bit easier.
Second part would be designing the keyboard. This is bit tricky as it needs to be designed with the UI and changing the physical layout would be fairly tedious. So you wouldn't want too many iterations of it. Good thing about the keyboard is that it could be made fairly independent of the rest of the system, and be pluggable via eg USB to PC.
And lastly would be putting everything together and on actual hardware. I'm a software guy, so I'm probably underestimating the work involved here. But imho it seems like getting basic prototype up and running hardware-wise shouldn't present too much trouble.
Of course once you reach the stage that you have mostly working prototype, then the real work begins. Designing a solid case, miniaturizing everything, optimizing power efficiency, and getting it to run on battery for a reasonable time. And at the same time honing the software to perfection.
For the hardware I have been thinking following configuration:
* 4-5" 16:9 display. Using widescreen might seem odd at first, but the idea would be that you could have two panes side-by-side to boost productivity.
* BeagleBone as the main board. Simply seems most suitable for the task. Small size, decent enough performance, good Linux support etc.
* Form-factor blatantly ripped off HP 200LX , one of the major inspirations for the whole project. I'm quite convinced that clamshell is the superior design for this kind of stuff.
* Case could be milled from aluminium. For prototypes I don't think the cost would be prohibitive compared to alternatives. And I'm not all that convinced of 3D-printed structural parts.
* There could be some kind of expansion port which has some GPIO pins and maybe ADC/DAC or something routed to it. That would allow the system to be used for data acquisition, which could make eg physics and electronics classes more interesting.
And for the software I'd have sage under ipython, and then a custom front-end for that. The front-end would most likely not be web-based like the current ones for various reasons. I had also some ideas about running everything under a tiling WM for multiple workspaces and some other stuff but that's fairly sketchy at this point.
 http://www.retroisle.com/others/hp95lx/Pictures/images_95lx/... That's a x86 computer running MS-DOS fitting on your palm. From the mid-90's. I think we can do something at least as cool today.
Right, I am assuming a soft keyboard as part of the UI here. But yes, eventually hardware keyboard is important.
> I'd (re)write the whole UI, designing it from scratch to suit small screen
This only needs to change if your screen is small -- otherwise the sage notebook seems reasonable at tablet sizes... but yes, if you're aiming at the Ti hand-held form factor it must be changed.
> I'm a _______, so I'm probably underestimating the work involved here
C'est la vie
> That's a x86 computer running MS-DOS fitting on your palm. From the mid-90's. I think we can do something at least as cool today.
Definitely. This project sounds awesome (and also a bunch of work) -- good luck!
TI calculators are awesome! Definitely feeling some nostalgia for coding in Z80 assembly language.
HN readers may be interested in efforts to unlock the OS on more recent models  .