I am color neutral, but not shape neutral. I always start the cross on the thinnest layer, as your eyes can (generally) recognize smaller shapes rather than bigger ones. Plus, that means that the OLL and PLL will stand out more, since they are bigger. For the OLL, you can see that the shapes sort of "look" like the corresponding color OLL. The no edges complete/all corners complete OLL still looks like a cage to me, the one that can be generated by F (U R U' R')2 F' still looks like two hammers. This is the best I can say without actually getting it out. But remember, there is no better way to improve times than with the P word.