The hash chain from chapter 1 says what a valid sequence looks like. It says nothing about who gets to add the next block, or what happens when two people try at once — which they will, because the network is not instantaneous. That is what a consensus mechanism decides.

Bitcoin's approach. To propose a block you must find a number which, mixed into the block, makes its hash fall below a target. There is no clever way to find it — you guess, trillions of times. The target is adjusted so that the entire network together takes about ten minutes. Finding it is enormously expensive; checking someone else's answer is instant.

That asymmetry is the security. To rewrite history you would have to re-win that lottery for every block since the one you want to change, faster than everyone else is winning new ones — which means controlling more than half the world's mining hardware. This is what chapter 1 meant by rewriting being expensive: not the hashing, the re-winning.

The cost is not metaphorical. Cambridge's 2025 study puts Bitcoin mining at roughly 138 terawatt-hours a year — about 0.5% of world electricity. Just over half is self-reported as renewable or nuclear by the miners surveyed, which is worth reading as a self-report rather than a measurement.

Ethereum's approach since 2022. Instead of burning electricity, a validator locks up a large sum of the currency as collateral and is periodically chosen to propose or vote on blocks. Provable cheating destroys part of that collateral — called slashing. Security moves from a physical cost to an economic one: attacking the chain means putting your own capital at guaranteed risk.

Finality is the point past which a transaction cannot be reversed. Bitcoin never quite reaches it — each new block only makes reversal exponentially less likely, which is why the convention is to wait for six blocks, about an hour. Ethereum gets closer to the real thing: a block is finalised after about 12.8 minutes, after which reversing it would require destroying more than a third of all staked funds.

A layer 2, in the chart above, is a separate faster chain that periodically posts a compressed summary of its activity back to the main one — borrowing the main chain's security while doing the work elsewhere. It is the main answer the field has to the throughput problem, and it works, and it still lands two orders of magnitude short of UPI.