You can measure something -- without actually doing the measurement.
You just threaten to do the measurement. The answer then appears.
It smells like a variant of 'spooky action at a distance'.
That's my summary of counterfactual computing. It will do for me until the Scientific American article comes out, which I will read and fail to comprehend.
This all feels as though we're hacking the infrastructure upon which our "program" runs.
If this stuff turns out to actually have real world implications the next forty years will be even less predictable than expected.
PS. Cosmic Variance has a serious attempt at an explanation.