The proposed approach seems biased towards short-term impact because of its simple evaluability. It is unclear, for instance, what the long-term impact of AMF or related charities will be. If we make "proven" refer to long-term impact, no cause would fulfil the requirements and it would come down to evaluating the expected utility (long-term) of the "promising" causes.

Granted, short-term impact is impact, too. But those who accept the arguments that the far future likely dominates by many many orders of magnitude would need to be very certain about long-term impact being virtually impossible to assess at this point. Maybe this case can be made (and Peter Hurford certainly makes a good attempt in his previous post), but I am not yet convinced.