Discover more from Something Interesting
Isn’t it better to burn money than coal?
Burning money is a dramatic gesture, but it won't secure a blockchain.
In this issue:
Isn’t it better to burn money than coal? (reader submitted)
Why are GPUs better at mining than CPUs? (reader submitted)
Would Proof-of-Stake be more efficient if staked capital could move freely? (reader submitted)
Isn’t it better to burn money than coal?
"I grant you that the costs of proof-of-stake are unquantifiable, but prima facie I would say that if I decide to burn a stack of dollar bills, then 99.9% of the costs are being born by me, whereas if I'm in Beijing and I'm burning a shit ton of coal for gov't subsidized energy, then likely society is bearing a lot of the cost." - IS
A really common response that I got for Proof of Stake will not save us was some variation on the idea that burning money didn’t hurt anyone because money isn’t a real resource. To be honest burning money is not a particularly good metaphor for what proof-of-stake is doing. Validators in a PoS protocol are not burning money, they are investing capital in exchange for a yield. Burning money is a kind of indirect donation to everyone. Being paid to put your money to work validating a PoS network is not. The whole idea is that you eventually get more money back.
More interesting though is the idea that society is bearing a cost as a result of China’s coal subsidies. I’m inclined to agree - I’m on record that I think we should stop burning coal. But the Chinese government would tell you the economic benefits of subsidized coal outweigh the environmental costs. That’s why I kept discussion of externalities to a subsection near the end and focused the general essay on efficiency - arguments about externalities are ultimately policy choices - how a society should balance trade-offs. Reasonable people can disagree about them.
I do think we should focus on taxing or subsidizing externalities as directly as possible, however. If we as a society decide Bitcoin mining is intrinsically distasteful and must be taxed so be it but if we think Bitcoin is bad because coal power is bad we should tax coal power. The more we can incorporate our beliefs about externalities into prices the more the market will naturally solve for efficiency on its own.
Why are GPUs better at mining than CPUs?
"Why are GPUs [graphics cards] better at mining than CPUs [computer processors]?" - BB
The Bitcoin network is designed to make sure that no matter how many people are mining new blocks are only discovered about once every ten minutes. To enforce that the network uses a parameter called difficulty. When too many blocks are found the difficulty is raised, when too few are found the difficulty is lowered.
What miners are actually doing is gathering up all of the valid transactions they have seen and assembling them into a data structure known as a block. To be a valid block it must hash1 to a certain number of leading zeros - the higher the difficulty the more zeroes are required. To find the next block in other words miners are trying to find a block that has the right hash value - so to do that they take the block and add a meaningless additional number called the nonce. Since the nonce doesn’t mean anything it can be changed, which will change the hash value of the block.
So what miners are actually doing is hashing the same block over and over again with different nonce values trying to find a valid hash with enough leading zeroes to satisfy the current difficulty. It turns out that swapping out a nonce and recalculating the hash is conceptually very similar to swapping out a pixel and raytracing it, which is what GPUs were optimized to do. GPUs are less flexible than CPUs but can parallelize the tasks they can perform more efficiently. Modern Bitcoin mining equipment is made up of ASICs2 that are a bit like hyper-evolved GPUs - even better at Bitcoin mining but no longer useful for anything else.
Would Proof-of-Stake be more efficient if staked capital could move freely?
"Let's say we get to a world in which ETH staking is using like 0.01% of bitcoin's electricity, never gets hacked and staked ETH is fully transferable. Would you then agree that proof of stake is a more efficient way to get security for blockchains?" - @CryptoDiplo
In Proof of Stake will not save us I talked about how the primary cost of PoS systems is that they lock up capital. Because that capital is locked up it can’t be used to build factories or research vaccines or buy me a pony or do any of the other pleasing things that capital can do. But what if there was a derivative token called stETH that represented staked ETH but was not locked, so you could stake ETH and still be able to spend it? Wouldn’t that reduce the opportunity cost of staking and therefore the social cost of proof-of-stake?
Unfortunately we can’t eat our cake and have it, too. If you spend your stETH you are no longer staking - the person you gave your stETH to is. You can’t stake your capital and spend it at the same time. Someone has to be willing to hold stETH in order for it to retain any value. If everyone is trying to sell their stETH it will be worthless.
That’s not to say stETH wouldn’t be useful - PoS protocols as we know them today have minimum quantities and lengths of time required by the protocol to stake. Derivative tokens would let investors enter and exit staking investments in smaller quantities and for shorter lengths of time, which would enable more investors to participate in the market. On the Ethereum beacon chain the minimum quantity is 32 ETH (~$75k at time of writing) and the minimum length is indefinite (since they don’t yet support withdrawing) so allowing investors to bypass those thresholds would be pretty significant. But making it easier for investors to put their capital into staking is not enough - you still need to motivate them to want to.
Investors want their money to earn more money, so they invest in whatever they think will earn the highest rate of return (for a given level of risk). To convince investors to stake ETH (or to hold stETH) the protocol needs to pay a rate of return at least as high as the next best thing investors could buy - and whatever they didn’t buy is the thing that society gave up in exchange for PoS security. That is the factory unbuilt, the vaccine unresearched, the pony you never gave me.
Transforming staked ETH into a transferable derivative like stETH would lower the costs of staking and help make the market for staking more liquid. More investors would stake, but by deciding to stake those investors are forgoing other things that society would benefit from. So fully transferable staking would improve proof-of-stake security (by attracting more capital to defend the network) but the total social cost of PoS would increase (because less capital would be available to do other things).
Other things happening right now:
The original Proof of Stake will not save us post generated a lot of interesting downstream discussion on Twitter. I encourage you to check it out:
Presented without comment:
Hash functions are math functions that transform any body of input into a number. They are one-way functions, which means it is easy to know what number corresponds to a certain input but impossible to know what input corresponds with a certain number. So if you want to find something that hashes to a specific value, the only way to do that is to guess and check until you get lucky and find something that works.
Application specific integrated circuits - basically special purpose custom made Bitcoin mining circuits. Like GPUs but optimized directly for mining instead of graphics.