How blockchains actually work

You do not need to understand blockchains to use crypto safely, in the same way you do not need to understand TCP/IP to send an email. But a working mental model makes you considerably harder to deceive, which is reason enough.

What is actually in a block

A block contains three things: a batch of transactions, a timestamp, and a cryptographic fingerprint of the previous block. That third item is what makes it a chain.

The fingerprint is produced by a hash function. Feed it any data and it returns a fixed-length string. The useful properties are that the same input always gives the same output, any change to the input — even one character — gives a completely different output, and you cannot work backwards from the output to the input.

Why tampering fails

Suppose you wanted to alter a transaction in block 500 of a chain that is now 900 blocks long. Changing it changes block 500’s hash. But block 501 contains the old hash of block 500, so block 501 is now invalid. Fixing that changes block 501’s hash, invalidating 502, and so on through to block 900.

So you would have to redo every block from 500 onwards — while the rest of the network continues extending the honest chain, which you must also outpace. On a large network this is not merely difficult; it costs more than any plausible gain.

How the network agrees

The hard problem a blockchain solves is agreement. Thousands of anonymous computers, some of which may be hostile, must converge on one version of history with no one in charge. The procedure for doing this is called consensus.

Proof of work

Proof of work makes proposing a block expensive. Participants — miners — race to find a number that, combined with the block’s contents, produces a hash below a target value. There is no shortcut; you guess, trillions of times per second. Whoever finds it first proposes the block and receives newly created coins.

The cost is real electricity and hardware, which is exactly the point: an attacker cannot fake it by creating more identities. Bitcoin uses this. The obvious criticism is the energy consumption, which is inherent rather than incidental — the security is the expenditure.

Proof of stake

Proof of stake replaces computation with collateral. Validators lock up coins, the protocol selects them to propose and attest to blocks, and misbehaviour destroys part of their stake. Attacking the network means acquiring an enormous stake and deliberately forfeiting it.

Ethereum switched to this in 2022, cutting energy use by over 99%. The criticism is that influence follows existing holdings, which may concentrate control among large holders and staking services.

Confirmations, and why they matter

When your transaction enters a block it has one confirmation. Each subsequent block adds another. More confirmations means more work an attacker would need to undo it, which is why exchanges wait for several before crediting a deposit. It is not bureaucracy; it is the probability of reversal falling toward zero.

“Decentralised” is a spectrum

It is tempting to treat decentralisation as a yes/no property. It is not. A network can be decentralised in who runs nodes but centralised in who writes the software, or in which few mining pools produce most blocks, or in who controls a protocol’s admin keys.

When a project describes itself as decentralised, the useful question is: decentralised in which respect, and who could still change or stop this? Many things marketed as decentralised have a small number of people who could halt them tomorrow.

What to read next

Next in this path: why does crypto have value? — which is a harder question than it first appears.