Of course, if you are intrigued and would like to pursue some of the following ideas yourself, just go ahead ― that's what they are here for. :)
This sort of machine could actually be more than just a toy or a curiosity, and prove useful for those tedious jobs when one needs to burn a staggering amount of CDs/DVDs. Its task list would be quite clear: grab a CD from a precisely located pile, drop it onto the drive tray, start the burning process somehow, wait for the recording to finish, grab the CD again, drop it onto a second pile, and repeat the process as long as necessary. NXT is, of course, best suited for this type of work, as it can communicate nicely with a computer.
The most difficult component of the multiplicator, at least within constraints set by the Lego parts, is obviously the CD grabbing mechanism. It needs to be able to pick up only the topmost CD from a pile that may not be perfectly aligned, yet be sufficiently accurate to place it onto the drive tray. I think it is possible, though far from simple.
There are many commercially available CD/DVD replicator systems, but they are usually too expensive, complicated and unreliable ― I've tested a couple of commercially available models and was largely unimpressed.
• Ferrari 348tb, Technic, 1:10
A Technic 348 would certainly, just like a real F348, need to have a mid-mounted V8 engine, a 5-speed gearbox, full independent suspension, steering, and openable doors in the least. If the construction allows, hidden headlights, adjustable seats, openable bonnet and boot would be nice bonuses, too. Or instead of them, a bit of PF.
There's no question about this MOC being possible, as one can easily find myriads of Technic sportscars on a similar scale and with similar functionality, but it isn't easy either, due to its high part density and high integration requirements.
• Heavyweight autofeeding scanner, Technic & NXT
The purpose of this project is not just curiosity, as I've got tons of stuff I'd need to scan, and I'll give a try with NXT before buying an off-the-shelf ADF scanner.
It should support at most the A4 format, and be able to hold at least approximately 200 pages, but 500 or even more would be nice. It would certainly be camera-based, with the separate sheets passing on sort of a belt underneath a suspended camera being triggered during every pass.
There's no doubt these components can be easily done, but what worries me is the feeder itself. This is a critical and difficult part even on commercial scanners, and preventing double-feeds (and detecting them if they happen) will be a massive problem. Perhaps constantly-rolling cascade of rubber wheels descending on page after page would take care of fetching the sheets, and an accurately trimmed pressure sensor could take care of double feeding ― it would measure the stiffness of the passing sheet in a curve, which would, of course, be significantly higher if two sheets are fed. A light sensor would make sure that each page passes through the process before the next one is fed. (Though, it may be also used to detect double feeding with a lamp on the other side of the paper...)
As there's much of manual work and some motors would run for a long time, it will probably require power supply directly from the outlet ― something along the lines Philo suggests on his website.
• Superheavy explorer spaceship, Classic Space, Minifig scale
It would need to have... well, everything, probably spread across several "floors", accessible in some clever way (Technic beams and axles...?). Specifically, the minimum: command bridge, crew quarters, navigation & communicatoon room, medical & bioanalysis quarters, storage rooms, at least one large and one small auxiliary airlock, scientific & engineering labs, umbilical arm to other spaceships, large fusion engine section with hydrogen tanks, shield generators and specimen containment. Then, a hangar with a small scout spaceship, two escape capsules and a medium-sized transporter, and a garage from which a few small buggies, expedition equipment, tracked mobile lab, mining vehicles and a mobile communication station can be deployed while surfaced.
Taking all into account, it should be at least 100-120 studs long, 60 wide and perhaps 25-30 high, and of roughly conical shape to cope with exploring the planets with an atmosphere. Probably with lots of attention to decoration and in blue-grey-black-white colour scheme to follow the Classic Space tradition. It will need to have a massive internal skeleton for rigidity (multiply reinforced Technic beams, I guess), and a few practical handles to pick it up, cleverly disguised as engine parts or landing skids. The crew of 30-40 people should suffice, with a few androids thrown in for variety. Some large functions, such as opening the garages, hangar, extending antennas etc. should be motorized, and there will be plenty of lights. Of course, this type of craft is doable, but on this scale and weight, it will require extensive structural and centre of gravity planning.
• CNC machine Gen II, Technic & NXT
Having failed the first iteration of an NXT-powered CNC machine featuring an off-the-shelf electric drill, I guess I have to attempt a Generation II. This time, with more prominent gearing down to provide sufficient power, stronger reinforcements, higher rotation speed of the drill and the sharper drill.
The ultimate goal is to carve materials such as soft wood, with the operational dimensions of approximately 10x10x5 cm, with precision no worse than half a millimeter. Speed is not of paramount importance.
Due to backlash involved between the gears, the amount of required down-gearing and the forces involved, I'm not entirely sure this is doable with LEGO, at least without insane amounts of bricks and reinforcements. A strong and reliable cradle for an electric drill is in itself a challenge already.
• Watermill/Water turbine, Technic
Obtaining mechanical power from a water flow is, despite lots of heavy difficulties, an attractive idea. Obviously, in the form of a watermill, it would need to use external watertight piping (beside many other its virtues, LEGO is next to impossible to build watertight), and is quite easy to build. It may power a cute little medieval-time minifig-scale mill ― if the wheel is large enough (say, 16 studs radius), it should easily provide sufficient torque and power.
However, a submerged turbine, while having no problems with watertightness as long as there are no electric parts, would be more of a challenge. The propellers would preferably have to be ducted as tightly as possible, no small feat with LEGO, and with the minimum possible detractions from the smooth surface that would reduce obtained power. But even with the most effective propellers and ducting, the torque on the output would likely be very small, requiring a clever transmission and gearbox gearing down with least possible resistance. So, the point is not merely getting mechanical power from the water flow (for that, just a simple submerged propeller would suffice!), but finding some optimal values and dimensions to get good efficiency.
• Floor plan surveyor robot, NXT
The chassis with its steering and power, and the log file analyser are the least troublesome part of this project. One can expect much more challenge wisely placing and adjusting the wall sensors and creating logic for executing accurate turns along the corners and edges (they need not be 90° always!). The vehicle would probably have to be medium-sized: large enough to stably drive over thresholds, but small enough to be manueverable. Very small turning radius is a must, probably implying four wheel steering and a wide control rod range. If there are some sensors left, it may be wise to add one small pair of spring-loaded "toucher" wheels in front of the vehicle as a precaution against it falling down the staircase.
It's certainly not impossible, but there's lot of programming, and this could be vastly more difficult project than it may seem at a first glance.
• The list goes on...