What Is Blockchain and How Does It Work?

A blockchain is a shared record of transactions that many independent computers hold copies of, arranged in batches called blocks, with each block cryptographically bound to the one before it. No single operator holds the master copy on a public network, and altering an old entry would require producing every later block again, at a cost the protocol deliberately makes high.

What is blockchain, and what problem does it address?

Blockchain is a method of keeping one shared record among parties who have no common administrator and no particular reason to trust one another. Every participant holds the same list of entries, and the rules for adding to that list are enforced by software.

Digital records are easy to copy, so an ordinary electronic payment message could in principle be sent twice. The usual answer to that problem is an institution — a bank, a registry, a clearing house — that keeps the authoritative ledger and decides which entries count. A blockchain replaces the single authoritative ledger with many identical copies and a public procedure for agreeing on what is added next.

The Bitcoin white paper, published in October 2008 under the name Satoshi Nakamoto, set out the first working description of such a procedure, and the first Bitcoin block was produced in January 2009. The same structure has since been applied to other kinds of record, including supply-chain documents, land registries and certificates of attendance. A blockchain is a record-keeping structure rather than a company, a product or a single network.

What does a block actually contain?

A block contains a batch of transactions together with a small header that identifies the block, links it to the block before it, and summarises the transactions the block carries.

A typical block header holds the following fields.

• The hash of the previous block, which is the link that makes the chain a chain.
• A timestamp recorded by the participant that produced the block.
• A single summary hash, often called a Merkle root, that condenses every transaction in the block into one value.
• Protocol fields such as a version number, a difficulty target or slot number, and, on networks that use proof of work, a counter that producers vary while searching for a valid header.

Below the header sits the transaction list itself, in a fixed order. Order matters: two transactions that spend the same funds cannot both be valid, so the position of an entry decides which one the network accepts. Blocks are numbered in sequence, and the very first block of a network is called the genesis block. Block sizes and intervals are set by each protocol; Bitcoin aims for one block roughly every ten minutes on average, whilst Ethereum divides time into twelve-second slots.

How does blockchain work when a new record is added?

A new record is added when a transaction is broadcast to the network, checked against the protocol rules by many computers, gathered into a candidate block by one participant, and then accepted by the rest as the next link in the chain.

1. A transaction is signed with a private key and sent to nearby nodes, which pass the transaction on to their own neighbours.
2. Each node checks the transaction against the rules: the signature must match, the funds must exist, and the format must be well formed. Entries that fail are dropped rather than reported.
3. Transactions that pass wait in a pending pool held separately by each node.
4. One participant assembles a candidate block from the pending pool and publishes the candidate to the network.
5. Other nodes re-run every check on the candidate. A block that satisfies the rules is appended to the local copy, and subsequent blocks are built on top of the accepted block.

Acceptance is a matter of degree rather than a single moment. Each additional block built on top of an entry increases the amount of work or economic commitment that reversing the entry would require, which is why entries are described as having a number of confirmations. Ethereum adds a separate procedure that marks blocks as final after roughly thirteen minutes, whereas Bitcoin treats settlement as increasingly firm without ever declaring a formal end point.

What is a hash, and why does it matter?

A hash is the fixed-length output of a mathematical function that converts an input of any size into a short string acting as a fingerprint of that exact input.

Three properties of the hash functions used in blockchains do the practical work. A hash function is deterministic, so the same input always produces the same output. A hash function shows an avalanche effect, so changing one character of the input produces an entirely unrelated output. And a hash function is hard to run backwards: no practical method is known for finding an input that produces a chosen output, or for finding two different inputs that produce the same one. The SHA-256 function, used by Bitcoin, always produces 256 bits, written as 64 hexadecimal characters, whether the input is a single letter or an entire library.

Hashing matters because each block header contains the hash of the previous header. Editing a transaction in an old block changes that block’s hash, which breaks the link stored in the next block, which changes the next hash, and so on to the tip of the chain. Anyone holding a copy can re-compute the chain and see the mismatch. The arrangement does not make a record impossible to change; the arrangement makes retroactive alteration detectable and computationally expensive, because every later block would have to be produced again.

What does distributed mean, and who runs the network?

Distributed means that the record is stored and checked by many independent computers, called nodes, each keeping its own copy and each applying the same rules without permission from a central operator.

Nodes are not all alike. A full node stores the history and checks every rule for itself. A block-producing node also assembles candidate blocks. A light client stores only headers and asks full nodes for the rest, which suits a phone or a browser. On a permissionless network, anyone with ordinary hardware, storage and a stable connection can run a full node, and no registration or approval is required. On a permissioned network — often used between named companies in logistics or finance — membership is restricted, and the participants are known to one another by contract.

Ownership of the network is therefore diffuse: individuals, businesses, universities and research groups all run nodes for their own reasons. Diffusion has limits worth stating plainly. Block production on large public networks tends to concentrate among a smaller number of well-resourced operators, and the storage needed to keep a full history grows every year, which gradually raises the cost of participating.

How does a network of strangers agree on one history?

Consensus is the set of rules by which independent computers, unable to identify one another, converge on a single ordering of blocks and resolve the moments when two valid blocks appear at once.

Consensus has two parts. The first part is a shared rule book that every node applies identically, so an invalid block is rejected everywhere without discussion. The second part is a chain-selection rule that picks between competing valid chains.

Proof of work as a mechanism

Proof of work requires a block producer to hash a candidate header repeatedly, varying one field, until the output falls below a numerical target. Finding such an output takes an enormous number of attempts and real electricity, whilst checking the result takes one hash. Nodes follow the chain carrying the most accumulated work, so rewriting history means out-producing the rest of the network for the whole period concerned.

Proof of stake as a mechanism

Proof of stake replaces the physical cost with a forfeitable bond. Participants lock a quantity of the network’s native units, the protocol selects one of them to propose each block, and a randomly chosen committee of the remaining participants signs statements that the proposed block is valid. A participant that signs two conflicting blocks can have part of the locked amount destroyed by the protocol. Ethereum changed from proof of work to proof of stake in September 2022.

Temporary forks occur naturally when two producers publish at the same height; nodes keep whichever branch carries more accumulated work or attested weight, and the other branch is abandoned, usually within a block or two. Permissioned networks generally use classical agreement protocols among a known membership, tolerating a fixed fraction of faulty participants. Consensus establishes an ordering that everyone shares. Consensus does not establish that the ordered statements are true.

How does a blockchain differ from an ordinary database?

A blockchain differs from an ordinary database mainly in who is allowed to write, who can check what was written, and what changing an entry costs after the fact.

An ordinary database is faster, cheaper and simple to correct, and correctability is a virtue in almost every organisation: an error in a customer record should be fixable in seconds. A blockchain gives up that convenience in exchange for a history that outsiders can examine without trusting the operator. Most record-keeping problems are better served by an ordinary database, and the honest question for any project is whether several parties genuinely need to share one record whilst distrusting one another.

What can a blockchain not do?

A blockchain cannot make a statement true, cannot make its participants anonymous, cannot operate without human governance, and cannot undo a mistaken transaction once the transaction has been included and built upon.

• Accuracy of inputs sits outside the system. A certificate written to a chain shows that somebody submitted that certificate at a given time; the record says nothing about whether the underlying goods, qualifications or ownership ever existed.
• Pseudonymity is not anonymity. Addresses are not personal names, yet every entry is public and permanent, and patterns of use can connect an address to a person, particularly where a service outside the chain holds identity documents.
• Governance persists. People write the software, operate the nodes and argue about rule changes, and disagreements have produced permanent splits into separate networks with separate histories.
• Capacity is limited. Public networks process a modest number of entries per second by design, and storing large files directly on a chain is impractical and costly.
• Programs deployed to a chain are not correct by virtue of being there. Defects in such programs are preserved exactly as faithfully as everything else.

What do the main terms mean?

The main terms of blockchain record-keeping are set out below in plain definitions, each with the Japanese rendering that readers in Japan usually meet first.

Blockchain (ブロックチェーン)

An append-only record of transactions, grouped into blocks linked by hashes and replicated across independent computers.

Block (ブロック)

A batch of transactions plus a header that links the batch to the previous block and summarises its contents.

Hash (ハッシュ)

A fixed-length fingerprint of an input, computed by a function that is impractical to run backwards and that yields an unrelated-looking output when the input changes by even one character.

Node (ノード)

A computer that holds a copy of the record, applies the protocol rules and relays transactions and blocks to its neighbours.

Ledger (台帳)

The complete list of entries, in order, from the genesis block to the most recent accepted block.

Consensus (コンセンサス)

The procedure by which independent nodes converge on one ordering of blocks without a central authority.

Private key (秘密鍵)

A secret number that produces the signatures authorising transactions from one address, and that is never shared or transmitted.

Public key (公開鍵)

A number derived from a private key that allows anyone to confirm a signature without learning the secret behind it.

Fork (フォーク)

A moment when two valid continuations of a chain exist at once, either briefly by accident or permanently through a rule disagreement.

Where can I read more about wallet security?

Wallet security is treated in a separate article on Metastarter, linked in the paragraph below. A private key is a secret number, and whoever holds a copy of that number can produce the signatures that move whatever is recorded against the matching address. Readers who have followed the mechanics of blocks, hashes and consensus often want to know how ownership is held in ordinary use, because the private key is the point at which the technical structure meets a person rather than a protocol.

The companion article, a security guide to browser wallets for beginners, sets out what a key does, what a seed phrase represents, and which patterns of deception recur — written as description rather than instruction. Nothing in that article asks the reader to do anything with a wallet, and no specimen key or phrase appears in it. The editorial approach behind both articles, including how sources are chosen and what Metastarter declines to publish, is set out on the about page.