The drawing to the left shows my original idea from several years ago. I have always wanted to make a puzzle where parts of apparently separate Rubik's Cubes could be mixed together. This desire probably goes back to my Siamese Cubes in 1981 where people would always ask if the parts could mix. Unfortunately they couldn't. In addition I have always thought the ring or torus was a much under used shape for twisty puzzles.
On a few occasions I looked into making this puzzle and had various ideas for a mechanism. However in early 2011 I accepted that making it by hand would be far too difficult and I would need help if it was ever to get made. This admission is a lot for me since I am very possessive about my ideas and see it as a failure to ask for help.
In April 2011 I contacted Oskar van Deventer (master of the 3D printed puzzle) and we discussed my idea. I showed him my mechanism diagrams but Oskar dismissed them in a flash and immediately designed his own vastly superior one. He also suggested that we get rid of the ugly triangular join areas and replace them with neat torus sections. Great idea which I never imagined would be possible.
Oskar quickly did the CAD work and a Shapeways printing machine became the third member of the collaboration. Printing the whole puzzle has little interest to me so I only asked for enough parts to make a set of moulds from. Unfortunately the core is too complex for me to copy so each puzzle I make has a 3D printed core plus hand cast cubies.
During the collaboration there were two minor disagreements between Oskar and myself. One of these was the name. He felt 'Cubes on a Ring' would be a better name but I went for 4Cubes instead. The second disagreement was the colouring. Oskar suggested having a neat logical arrangement where all sides matched. This would have made a nice conventional looking puzzle but that is not what I wanted to achieve. I wanted the four cubes to appear as independent as possible. As if you have randomly taken four different solved Rubik's Cubes and joined them together with out worrying about matching colours.
This means there is no 100% correct solved state for this puzzle. As long as each individual cube is solved it does not matter how they relate to each other. You might think this makes the puzzle much easier but I was careful to colour it in a way that meant it isn't. Ignoring complete cube rotations and positionings, each corner and edge cubie has only one correct position on the puzzle. In addition my colouring allows for a fuller looking mix-up since all six colours can appear at all locations.
One of my concerns when we started this project was the movement. I could see that all four cubes would have to be perfectly sitting on the ring to allow for the large rotation moves. This meant the puzzle needed to have quite stiff movement so the cubes are not moving around freely when you pick it up. Even with stiff movement I discovered the best way to use the puzzle is on a flat surface when the large rotation moves are desired. The video shows this much better than I can explain it.
Having made the prototype I went on to make a few more. Unfortunately though there are quality issues with those and for some reason there appears to less clearance between the cubes. This results in clashes of pieces which damages stickers.
Update: The very day I revealed this I discovered that Sergey Abramenkov patented the first design in 2010. This is pure coincidence since I never saw his puzzle and I know he never saw mine.