You can get a pretty accurate measure of gdp for the currency in real time from the block chain. Even easier if transaction fees are fixed at a percentage of the value. Can't you use that to give us a formula to fix supply for a stable currency, macro economist.
If possible, this might be a good idea. As far as I'm aware though, the block chain will tell you at most nominal GDP, not real GDP, but NGDP targeting can be close to optimal under certain circumstances. The strongest arguments for NGDP targeting are based on wage rather than price rigidity, but no one that I know of is currently paid a bitcoin wage, so this may be one for the long-run.
If you were to do this, you'd tie the interest/demurrage rate to the deviation between NGDP and an exponential trend (say 2%). So when NGDP was high you'd increase future interest payments more than 1 for 1, for Taylor-principle type reasons.
However, I'm a bit sceptical about the claim that you can get even NGDP from the block chain. In order to arrive at NGDP, you need to know which transactions are associated with the exchange of goods, and which of these exchanges are intermediates for the production of other goods. So, for example if I have two wallets, and I send bitcoins from one to the other, as far as I'm aware, that transaction will appear in the blockchain just the same as one in which I sent someone some bitcoins and they sent me some chocolate. The latter should be counted in NGDP, the former should not. Similarly, if Bob buys cocoa beans with bitcoins from Alice (who harvested them), then Bob uses them to make chocolate, which he sells to Charlie for bitcoins, then if we count both transactions towards NGDP (Alice to Bob, Bob to Charlie), then we are double counting the cocoa beans. True NGDP in this case is just the value of the second transaction.
Thus extracting NGDP from the blockchain seems like an impossible undertaking. But perhaps I'm missing something.