I've kicked around the idea for a while of a baggable computer, in the sense that the components might fit into a satchel or messenger bag, but be connectable to a larger set of I/O devices (keyboard, mouse, display) as needed.
For the past year and a half I've been using the Onyx BOOX Max Lumi e-ink tablet. It's sized large enough at 13" to be a very useful display. As an Android device, it ... suffers many of the limitations of same (see: <https://news.ycombinator.com/item?id=32569871>).
But paired with a Real Computer, a keyboard, and a mouse or trackball, plus battery/power supply and maybe external disk storage ... it could be interesting. Those might fit into different pockets of a bag, though this would leave the problem of thermal management (cloth-swathed electronics have poor heat conduction characteristics).
Different components could be swapped out at will. And the device (or its storage) could be paired with a Real Desktop Computer (or work-in-process synched to same or some common storage pool) as needed.
The key constraints for a laptop itself are the fact that the keyboard and screen must accommodate a human's hands and eyes, and provisioning sufficient battery capacity. Busting a few of those constraints for more flexible offerings ... might shake up the concept a bit.
For modern generative computing, web browsing is typically the biggest load. I've used 10--15 y.o. (or older) hardware quite usefully for most of my principle tasks except Web access, which ... is kind of sad on the face of it. Even an older Android tablet runs scripts (generally awk or python) reasonably well. Runtime is about 1/10 the performance of a desktop, possibly worse, but for small working data sets, that's rarely an issue. E.g., I can run awk with effectively the same responsiveness on many thousands of rows of data in either environment. I did find that JSON parsing was notably slower (10--15s on the tablet vs. sub-second response on desktop) for typical workloads, though even that was an issue only in high-repetition development or content-search contexts.