I really don't have a to-do list, because I'm not really working on the project. Let me throw out a few ideas, though:
1. It'd be nice to use a proper parser (one that actually understands all of the mediawiki syntax, templates and lists etc). Best would be to use real existing code that does this, but change the rendering from html to mostly text-only, exceptions being: links, headings (for navigation, though the links generated by a toc template might obviate the need for this), maybe some basic formatting like bold italic and underline, and recognizable tables (wikipodia has an anchor in the article and the table itself as another screen). then:
2. rendering engine that supports these. the frontend is simple enough that it would be profitable to rewrite it with a good plan in mind, it's sort of a mess.
3. Using a less hacked-up gzip might be nice, too, and reconsidering bzip2 I think would be worthwhile. As the entire archive uses about the same character set it might be worthwhile to do a first huffman coding pass over the whole text or something, to factor out common things. Testing of course would show if this actually helps...
4. the paged scrolling thing the guy keeps posting about. I prefer line-by-line but I understand how others might like full pages and it'd be an easy bit of configuration (any number of lines, even). other configuration options would include backlight timing, save on exit, font selection (for those who want a pretty system font but the range of unifont for wikipedia articles).
5. a better article selection interface. traversing the tree as characters are entered, and a scroll of only valid characters (and completing the rest of the line for the first match) is a must. This will also have the bonus feature of allowing article selection for any language (though this will mean probably some work on the sort order).

A wishlist featue, which I think is nevertheless very important even with the prevalence of broadband, is a distribution system that can apply diffs (conceptually, that is) to keep the archives updated locally without excessive downloads.
I have given the project a lot of thought, but unfortunately I have not had the time to actually hammer out any code. I'd be thrilled to see any new work that you could contribute.