Blockchain is only potentially useful if there’s no single entity that can be trusted. If banks can’t even trust themselves to manage their own internal ledgers, they have much bigger problems to deal with.

Blockchain has been around as a technology for nearly two decades. If financial institutions thought it could help them you can bet they would be all-in on it by now. As it is, blockchain has no significant advantages over traditional financial ledger systems, so what incentive is there for them to use it.

It's not something new or cutting edge any more, just waiting for a bright spark to discover the technology and put it to use.

How is the blockchain different from a read only ( write only once to be specific) DB that follows ACID?

Why would you want the computational power of a bank system have anything to do with whether it's ledger is correct?

Yeah let’s use the computing power of an entire country to pay for a small coffee.

Absolute immutability is kind of a terrible property for a financial system though, cos it completely ignores the fact that mistakes and fraud happen and you need a way to forcefully recover funds other than "lol sucks to be you I guess".

The one actually genuinely useful application for this kind of technology that anyone has come up with is Certificate Transparency, but crypto people don't get excited about it cos it's not possible to make money from it.

the technology itself has its use cases.

Cool.

Name one successful example.

I mean, it's been, what, 15 years of hype? Surely there must be a successful deployment of a commercially viable and useful blockchain that isn't just a speculative cryptocurrency or derivative thereof, right?

Right?

There is nothing Crypto can do better than regular database. Immutability is not a desirable property.

Well yes but can you pump and dump git commits ?

Blockchain is pretty well defined.

Git doesn't have update rules that are only valid if signed by a particular private key.

The independent currency was still worth standard currency... So the ones already hoarding all of that existing make-believe-number money just bought up and schemed us out of new make-believe-number money.

How did we not see this coming? :(

Amazon has a database project called QLDB that uses blockchain for its audit log.

What do you call an audit log that's cryptographically verifiable? A blockchain.

Blockchain is often described as a solution in search for a problem. It’s a clever technology, but people don’t really know what it can be used for besides storing cryptocurrency transactions.

People have thought about storing other kinds of data in the blockchain, like health records, but no one can really point out to why this would be better than other solutions.

To achieve something similar with health records without blockchain, all that is needed is just a cryptographic signature. The hospital cryptographically signs a digital health document and email it to you. The hospital in turn stores it in some shared database accessible by other hospitals. Done.

If the health record is somehow lost from the shared database, then you got your own copy of it as backup. They can’t modify the health record either, because then it would diverge from your own copy.

The worst thing they can do is to add falsified health records without your approval, but that’s a problem with blockchain as well. Blockchain cannot verify that the input data is truthful (garbage in, garbage out).

The cryptographic signature step is a part of blockchain either way, so there’s no extra technical overhead in the non-blockchain way.

The cryptography has much simpler algebraic analogues - what we are looking for is a "one-way function". This means a mathematical symbol that only works on the left side of the equals. The simplest one is the remainder of a division. For example if I told you that I had a remainder of 5 after dividing by 20, you wouldn't know if the original numerator was 25, 45, 65, 85, and so on. This operator is called mod (modulus). Even if you don't know what value I started with, It's not hard to guess what possible numerators could be with modulus. That's where the cryptography comes into play: a cryptographic hash is designed so that it's practically impossible to guess the original numerator. We'll stick with the modulus for explanatory purposes, but imagine that you can't list off possible numerators like I did.

Now we can invent a puzzle for a computer to solve. We'll start off with the same values as before, but - again - we are disallowing easy guesses. This forces us to check 1 mod 20, 2 mod 20, 3 mod 20, 4 mod 20, 5 mod 20 and so on. Eventually we'll hit 25 mod 20 giving us the solution to X mod 20 = 25. Now you can go back to the person that gave you the puzzle and prove that you've done 25 steps of work to arrive at a solution (or have made a lucky 1/25 guess). This is called "proof of work". A cryptographic has consists of a certain number of bits, such as 256 bits - this means a series of 1' s and 0's 256 long. The puzzle presented to the computer is "find the numerator that results in the first 50 bits being zero" (the more bits are required to be zero, the longer it will take to find the answer). Because of the incredibly slim chance of guessing the correct numerator, it doesn't really matter if the computer counts up (like we did with modulus) or guesses. So, in practice, everybody trying to find the solution starts at a random number and starts counting, or trying other random numbers, until someone wins the jackpot. It's basically a lottery, but the correct numbers have to be discovered instead of being dropped out of a glass ball at the end of the week. Once a computer finds a solution, everybody else playing the game can check their numerator as [probabilistic] proof that they have done work.

Now we can use this lottery to create a blockchain. We start with 5 things: a globally agreed on solution we are looking for (789), an initial block (which is just a number - lets say 12345), Bob's account #5 of $100, and Sally's account #6 of $200, and a huge amount of players of the above game. Sally wants to transfer $20 to Bob, so she says to all the players: "I'm #6 and want to give #5 $20. There's a $1 prize for finding a new block for me." All the players make a new denominator, by placing the numbers next to eachother - so 12345 6 200 5 100 20 1 - or just 1234562005100201. All the players start trying to find the number that will result in 789. Eventually someone finds 1234562005100990 after a lot of work/guesses. Everybody checks their work 1234562005100990 mod 1234562005100201 = 789. The winning player receives their prize, and now everybody has a new block to start from: 1234562005100201 1234562005100990. Next time someone wants to send some money they will use 12345620051002011234562005100990 as the initial block instead of 12345. Hence, we have set up a chain starting with:

12345 -> 12345620051002011234562005100990 -> ...

There's your block...chain. Anybody can independently verify that the work has been done by checking the answers. It's incredibly elegant but, as we've seen, incredibly destructive.