The ideal would be useful work that is a public good - no one can profit from it individually (other than by using it as a PoW), but the world benefits from it in some way.
Like Protein folding, Search for ExtraTerestrials, Asteroids, etc... there are a lot of them, some of which structured properly could provide public benefit.
I'd like to suggest "AlienCoin": Proof of work consists of processing SETI@Home blocks, and mining rewards are paid out to whoever finds an alien civilization.
Satoshi style PoW algorithms are intended to waste as much energy as possible.
PoW mining just raises the capital cost to control the network without ever improving the speed or functionality. It's a system that increases the waste as time progresses.
I wonder about some tasks that people already do for charity with their computers...like protein folding, finding super huge prime numbers. I don't know enough about the algorithms involved though to assert they satisfy the POW requirements. But some infrastructure already exists to distribute these problems.
It was more of an off-handed comment, but I suppose you could rig some sort of carbon sensor that measures X amount of carbon pulled from the atmosphere. Of course any sort of sensor to the physical world can be rigged to produce a false signal.
You and I were thinking on similar lines. Since it's tech, my idea was the money went to FOSS nonprofits that focus on infrastructure like OS's, servers, crypto libraries, and so on. Even the GetRichCoin types building on that kind of foundation might do some incidental good for tech scene even if their plans don't work out.
