Make rewriting expensive.

First, the rules still apply.

A payment must satisfy conditions such as proper authorization and not reusing funds already spent in that history. Blocks must also follow rules about their structure and allowed new issuance.

Computers called full nodes check these rules. A miner cannot add an unauthorized payment and make it acceptable by spending more electricity.

That distinction prevents a common misunderstanding. Work does not decide whether two plus two equals five, whether a stranger owns your key, or whether extra coins outside the issuance rules are permitted.

Start by removing invalid candidates. Only then ask which of the remaining histories has the most accumulated work.

A costly search, an easy check.

A miner gathers candidate transactions into a block and repeatedly changes small pieces of data, testing the result with a cryptographic hash function. You will open that mechanism in the lesson on mining. For now, think of a repeatable check whose successful result is difficult to find.

The block’s result must fall below a specified target. Most attempts fail. A miner with more hashing capacity can try more candidates per second and has a greater chance of finding the next valid result.

Once someone finds one, others can check the proof without repeating all the failed attempts. They also check the block’s transactions. If it passes, other miners can build on it.

The successful miner can claim the reward allowed by the rules, including the permitted subsidy and transaction fees. This gives miners an incentive to publish valid blocks others will accept.

Which valid branch should grow?

Sometimes two miners publish different valid next blocks close together. Different nodes may temporarily favor different branches.

Bitcoin follows the valid chain with the most cumulative proof of work. As more work is added to one branch, nodes can converge on it. This is not a count of people, and it is not simply a count of blocks when their difficulties differ.

“Work” here is a measure derived from the block targets. The protocol does not read electricity meters. Machines differ in efficiency, and finding a block involves luck.

A transaction has its first confirmation when it is included in the current accepted chain. Further blocks build additional work on top of it. That generally makes replacing its history harder under the security assumptions.

An attacker has a moving target.

To replace a payment already in the chain, an attacker can try to build an alternative valid branch that spends those same funds differently. The attacker must catch up while other miners may keep extending the existing branch.

With less hashing capacity than the honest miners combined, success becomes less likely as the payment gains confirmations, under the model’s assumptions. Majority hash power changes the risk dramatically. Neither case turns an invalid signature into a valid one.

Mining hardware, electricity, and forgone rewards make attacks costly. The design aims to reward participation more than disruption. It does not prove every possible attack is permanently unprofitable, or that every payment is certain and irreversible immediately.

The result is a system with explicit rules and economic barriers to replacing history. It is more precise—and more useful—to understand those barriers than to call the record impossible to change.

The idea to keep

Check validity first. Compare accumulated work second. Additional confirmations make replacing an accepted payment more difficult under the system’s assumptions.

The rules prevent forbidden entries; proof of work helps order permitted ones. You need both parts to understand Bitcoin.