I've been looking for an answer to this question for a while -- why not build a digital commodity currency, where useful computation is a commodity? (I've yet t

I've been looking for an answer to this question for a while -- why not build a digital commodity currency, where useful computation is a commodity? (I've yet to find a satisfying response to this). Several problems come to mind: * There's not enough computation. With BTC-type currencies, the value of the money is far higher than the cost of the backing computation (look up bitcoin mining calculations). If you have a computational commodity currency, they should be closely linked (by definition!). And a home PC can support a trivial amount of computation ($100's at most) -- not enough to be useful. * Not enough demand for (pure, parallelizable) computation. * No intrinsic way to "store" computation. You can record the existence of computation, but if this done by a distributed P2P system like bitcoin, the commodity-backing property is lost. A file verifying that a computation was performed is not itself computation: it doesn't solve any problems, it has no intrinsic value. Of course a legal private currency, backed by banks and courts, could keep track of such debts and make them enforceable. * Security problem: you would need to run arbitrary machine code on a home PC, if this were to have any useful value. This is dangerous and unsolved. Is there any way to run, e.g., untrusted CUDA code in a sandbox? (I believe no) * Accounting. There's no pure "unit" of computation: you would have many types of computational problems, being assigned to many different types of hardware. In the real world, I think you wouldn't end up with a single currency so much as a market, with many currencies and many fluctuating exchange rates. If you could extract a currency from this, it would be as a currency basket (5% protein folding, 5% linear algebra...) So I suspect this wouldn't work well today. Maybe in the future.