Almost everyone who uses Bitcoin asks someone else for their balance. You open your wallet, it queries a server, the server responds with a number, and you trust it. The vast majority of the time, the number is correct. But notice the verb: trust. There is an alternative, and it fits on a regular computer on your desk.
A full node downloads all the blocks since January 2009 and checks each one independently. It verifies the signature of every input of every transaction, ensures no coin is spent twice, recalculates the merkle root of each block, redoes the hash of each header against the target of that time, and confirms that the reward paid to each miner was exactly as allowed by the issuance rule. That's over a billion transactions audited one by one, resulting in a set of unspent outputs it built on its own, without asking anyone if it's correct.
The initial synchronization is the costly part. It's a few hundred gigabytes to download and verify, taking from a day to several days, depending on the machine and internet. After that, the node only keeps up with what's incoming — a few megabytes every ten minutes, with energy consumption similar to a light bulb.
One configuration deserves explanation because it's often misunderstood. The program comes with an optimization that skips signature verification on very old blocks, below a marked point in the code. It's not blind trust: that point is the hash of a block anyone can verify from twenty independent sources, and anyone who wants to disable the optimization and verify signature by signature since 2009 can do so with a configuration line.
If the disk is small, there's pruning. The node downloads each block, verifies each block, and then deletes the old ones, keeping only the set of unspent outputs and recent blocks. The minimum fits in less than a gigabyte. Pruning is not trusting third parties: it's verifying and then discarding the proof.
In practice, what changes for you? Three things. The wallet starts asking your node, and the response no longer depends on a company's honesty. Your privacy improves significantly because you stop informing an external server of all the addresses you own — which, as the lesson on privacy will show, is the most common leak there is. And, most importantly, you start enforcing the rules: if tomorrow the majority of miners decide to create coins outside the planned issuance, your node rejects those blocks and continues on the chain that follows the rules you accepted. This is exactly what decided the 2017 dispute, as the module on what Bitcoin is has already explained.
It's worth being honest about what running a node is not. It doesn't make money, it's not mining, and it doesn't speed anything up. It doesn't protect you from losing your seed or falling for a scam. And most people will never run one, which is fine: the network needs running a node to be cheap and possible, not for everyone to run one. The value lies in the cost being low enough for the option to exist for anyone, anywhere, without asking for permission.
A hundred-dollar computer and a disk will do. It's the difference between having an opinion on Bitcoin's rules and being one of the places where they happen.
With the base layer understood from the inside, questions begin that it alone doesn't answer: how to pay for a coffee in a second, what changed with Taproot, why there was a block war, and what exactly the blockchain reveals about you. That's the subject of the next module.