Stuff that'd be cool to see...

Just a little random list of things that I imagine would be quite nice / interesting to see:

• An animated short film or even a full length feature film rendered without wireframes showing (and possibly no or few textures) - I'm sure that we've all seen those 3d model turnarounds where people show their wireframes for their models. Well, IMO these look quite beautiful as they are already (especially when these have really great topology), and even more so in motion. Hence, I think it would be really cool to see this once - who knows, it's possible that the producers of such a film could get away with a lot less, especially by reducing the need to have such highly detailed lighting + shading (though interesting lighting would also come into use). 
• A computer simulation capable of approximating timing+frequency+strength of earthquakes (i.e. "earthquake forecasts") that actually works - I know, it's probably wishful thinking from a resident in a quake-stricken city, but clearly we currently don't know enough about earthquakes yet to do anything close to this other than to say that there is a very vague probability for a large period of time (and then say that as with all probabilities, there are always anomalies, some larger than others - in effect saying, we don't really know what will happen... we aren't there yet despite years of research+funding). From a computer science point of view of this problem, it's entirely plausible to think that there must be actually some (simple, as always) set of underlying algorithmic processes, which, when combined with calibrations against historical earthquake sequences and general terrain information might yield a combination of Pseudo-Random Number Generator + seed which would be capable of quite accurately approximating said earthquake sequences for some region. Of course, the likelyhood of actually finding such combinations is probably quite low, especially given the difficulty in identifying the relevant variables that would lead to this, hence why this approach may have been abandoned (or outright rejected) in the past.
• Tool for posing characters using freehand sketching interfaces - it's still too early days yet to really say anything conclusively on this matter, but I'm aware of research over the past few years into such technology. How flexible it is in practice, or whether it will actually be that useful for animators in the field has yet to be seen once "widespread dissemination" of the technology has actually taken place. Wouldn't it be cool if Blender was that vehicle ;)