my reply from the other thread fits here too:

> this is just not true at all, there are massive leaps from algorithms, data, etc. every year. scale is one axis of many and you need to get them all correct.