A blockchain is a ledger that thousands of computers keep identical without trusting one another. New entries arrive in batches called blocks, every computer checks each batch against the same rules, and the batches are linked so that changing an old entry would be obvious and ruinously expensive. This guide builds the system from its parts, using Bitcoin as the main example and Ethereum as the contrast, then follows one bitcoin payment from wallet to sixth confirmation. Why no blockchain can maximize speed, security and decentralization at once is covered in the blockchain trilemma, and the crypto hub covers the assets that run on these networks.
What is inside a block?
A block is a header plus a list of transactions; Bitcoin's 80-byte header holds six fields, including the previous block's hash and a summary of every transaction.
| Header field | Size | What it does |
|---|---|---|
| Version | 4 bytes | Signals which rule upgrades the block supports |
| Previous block hash | 32 bytes | Links the block to its parent; change the parent and this no longer matches |
| Merkle root | 32 bytes | One hash that summarizes every transaction in the block |
| Time | 4 bytes | The miner's timestamp, which must be later than the median of the previous 11 blocks |
| Target (nBits) | 4 bytes | A compressed form of the threshold the block's hash must fall below |
| Nonce | 4 bytes | The number miners keep changing while searching for a valid hash |
The transaction list is limited by weight rather than count. Bitcoin allows 4 million weight units per block, which with today's mix of transactions works out to a few thousand payments; the 144 blocks up to the first one mined on November 2, 2026 averaged about 4,705 each. Ethereum measures blocks in gas instead, the computing work each transaction needs, and validators raised the limit to 60 million gas per block on November 25, 2025, days before the Fusaka upgrade.
What does a hash do in a blockchain?
A hash function turns any data into a fixed-length fingerprint, and the smallest edit to the data produces a completely different fingerprint.
Worked example: One character, a completely different SHA-256 hash
SHA-256, the function Bitcoin uses, always returns 256 bits, written as 64 hexadecimal characters.
- "Pay Alice 0.01 BTC" hashes to 8d83233da9a57ebdf515abcfe23c0313b6ce1b14ffe75300aa26f39967d9ee16
- "Pay Alice 0.02 BTC" hashes to 5abc5c830da6ef116b0a18b6cab2bbb968ede92fe1b50d857a614b26691000dd
- One digit changed, yet the outputs share no visible pattern, and the only way back from an output to its input is guessing.
Blockchains use this property twice. First, for chaining: each header contains the previous header's hash, so altering a transaction in an old block changes that block's hash, which no longer matches the copy stored in the next block, and so on up to the newest one. Every honest node would see the mismatch at once. Second, for proof of work: Bitcoin miners hash candidate headers until one comes out below the target, which shows up as a run of leading zeros. The first block's hash in 2009 began with 10 zeros and the hash of block 974,194, mined on November 2, 2026, with 20, a rough sign of how far difficulty has climbed, from 1 to about 148 trillion. The what is bitcoin guide explains how that target is set.
How do Merkle trees keep blocks compact?
A Merkle tree hashes transactions in pairs, then hashes those results in pairs, until one root hash remains that changes if any transaction changes.
With four transactions A, B, C and D, the tree hashes A with B and C with D, then hashes those two results together to get the root that goes in the header. To prove that C is in the block, you need only D's hash and the combined hash of A and B; anyone can recompute the root from those and compare it with the header. The proof grows with each doubling of the transaction count rather than with the count itself.
Worked example: How big is a Merkle proof for a real block?
- Block 974,194 holds 2,286 transactions.
- 2 to the 11th power is 2,048 and 2 to the 12th is 4,096, so the tree is 12 levels deep.
- A proof needs one 32-byte hash per level: 12 × 32 = 384 bytes.
- A block with 4,705 transactions, the average over that day's 144 blocks, needs 13 hashes, or 416 bytes.
- Every Bitcoin header since 2009 together takes about 78 MB (974,195 headers × 80 bytes), small enough for a phone.
This is the simplified payment verification idea in section 8 of the Bitcoin whitepaper: a light wallet keeps headers, asks a full node for the Merkle branch and checks that the payment sits under a header with plenty of work on top. It trusts that most of the mining power is honest, which is why the paper suggested that businesses receiving frequent payments run their own full nodes.
How do digital signatures and addresses prove ownership?
A wallet holds a private key, derives a public key and an address from it, and signs each payment so anyone can verify it without seeing the key.
Bitcoin and Ethereum both use the secp256k1 elliptic curve. A private key is a 256-bit random number, and multiplying a fixed point on the curve by it gives the public key, a step that is easy forward and impractical to reverse. A signature proves the signer knew the private key for that exact transaction, so changing the amount or the recipient afterward invalidates it. Bitcoin uses ECDSA signatures and, since the Taproot upgrade activated at block 709,632 on November 14, 2021, Schnorr signatures as well.
Addresses are shortened, error-checked forms of public keys or spending scripts. Native SegWit addresses start with bc1q, Taproot addresses with bc1p, and older ones with 1 or 3. An Ethereum address is the last 20 bytes of the Keccak-256 hash of the public key, written as 0x followed by 40 hexadecimal characters.
The two networks keep score differently. Bitcoin tracks unspent outputs, separate chunks of bitcoin each locked to an address, and a payment consumes whole outputs and sends any change back to the sender. Ethereum keeps a balance and a transaction counter for each account, so the same signed transaction cannot be replayed. Either way, whoever has the private key has the coins, and custodial vs non-custodial wallets explains who holds that key in each kind of wallet.
What happens in the mempool?
The mempool is each node's waiting room for valid transactions that are not yet in a block, and miners usually fill blocks from it by fee rate.
There is no single mempool. Each node keeps its own, so two block explorers can show slightly different queues. Bitcoin Core's defaults cap a node's mempool at 300 MB and drop a transaction that has waited 14 days, and since version 29.1, released on September 4, 2025, a node relays transactions paying as little as 0.1 sat/vB by default, down from 1 sat/vB.
Because block space is auctioned by fee rate, a stuck payment can be rebroadcast with a higher fee, a method called replace-by-fee that Bitcoin Core has accepted for any unconfirmed transaction by default since version 28.0.
Shortly after midnight UTC on November 2, 2026, the queue was long but cheap: about 49,600 transactions taking up 20.9 million virtual bytes, roughly 21 blocks' worth, yet offering only 0.03 BTC in fees combined. Ethereum's mempool works on the same principle, except that the protocol sets each block's base fee and burns it, and senders add a tip for the validator; gas fees explained covers that system.
How do proof of work and proof of stake differ?
Both decide who adds the next block and make cheating expensive, but proof of work spends electricity on hashing while proof of stake puts deposited coins at risk.
| Proof of work (Bitcoin) | Proof of stake (Ethereum) | |
|---|---|---|
| Who adds the next block | Whichever miner first finds a hash below the target | A validator picked at random for each 12-second slot |
| What it takes to join | Mining hardware and electricity | A deposit of at least 32 ETH and node software |
| Block timing | Random, averaging about 10 minutes | Fixed 12-second slots; 32 slots make a 6.4-minute epoch |
| Penalty for cheating | Electricity wasted on blocks the network rejects | Slashing, which destroys part of the validator's stake |
| What an attacker needs | More hashing power than the rest of the network | A third of all staked ether to stall finality; reversing a finalized block costs at least a third of the stake |
| Scale on November 2, 2026 | About 1,046 exahashes a second | About 44.3 million ETH staked by 815,282 active validators |
| When a block counts as final | Never absolutely; reversal odds fall with each block | After about two epochs, roughly 13 minutes, once checkpoints are finalized |
The 32 ETH deposit is a minimum; since the Pectra upgrade on May 7, 2025, a single validator can hold up to 2,048 ETH of effective balance. Ethereum's staking system is covered in what is ethereum and the risks of staking your own coins in what is staking, while bitcoin vs ethereum sets the two networks side by side on fees, supply and volatility.
When is a blockchain transaction final?
On Bitcoin a payment never becomes final in an absolute sense and simply gets harder to reverse with each block, while Ethereum finalizes blocks after about 13 minutes.
Each block added on top of a transaction is a confirmation. Section 11 of the Bitcoin whitepaper models an attacker racing to build a secret chain that leaves the payment out: with 10% of the hashing power, the attacker's chance of catching up is about 20.5% after one confirmation, 1.3% after three and 0.024% after six. The common advice to wait for six blocks comes from that math.
Reorganizations are not only theoretical. In March 2013 a database limit in older Bitcoin software made some nodes reject a valid block and split the chain, and BIP 50, the post-mortem, records at least one large double spend before miners reunited the network, made as an experiment rather than a theft. Short reorganizations of a block or two still happen when two miners find blocks almost simultaneously.
Ethereum adds explicit finality on top of its rule for choosing between forks. The first block of each epoch is a checkpoint, and when validators holding at least two-thirds of staked ether vote for a pair of checkpoints, the earlier one becomes finalized. Undoing a finalized block would mean at least a third of all staked ether being destroyed, which at 44.3 million ETH staked is more than 14.7 million ETH. If finality stalls for more than four epochs, an inactivity leak drains stake from non-voting validators until the rest can finalize again.
How does one bitcoin payment travel from wallet to confirmation?
Here is one ordinary payment, 0.01 BTC sent from a wallet holding a single 0.05 BTC deposit, followed from the moment it is signed to its sixth confirmation.
Worked example: 0.01 BTC from wallet to six confirmations
Assumptions: native SegWit addresses, a fee rate of 2 sat/vB, comfortably enough for the next block in early November 2026 when explorers recommended 1 sat/vB, and an assumed price of $80,000 per bitcoin for the dollar figures.
- Build. The wallet spends its one 0.05 BTC output and creates two new ones: 0.01 BTC to the recipient and change back to itself. One input and two outputs take about 10.5 + 68 + 2 × 31 = 140.5, rounded up to 141 virtual bytes.
- Price. The fee is 141 × 2 = 282 satoshis, about $0.23, so the change output is 0.05 − 0.01 − 0.00000282 = 0.03999718 BTC.
- Sign and send. The wallet signs with the private key and hands the transaction to a node, which checks the signature and that the 0.05 BTC output is unspent, then passes it to its peers.
- Wait. The payment sits in mempools with thousands of others. Blocks arrive at random, so the chance the next one comes within 10 minutes is 63%, and within 30 minutes 95%.
- First confirmation. A miner includes it because 2 sat/vB beats most of the queue, and the recipient's wallet now shows one confirmation.
- Six confirmations. Six blocks take 60 minutes on average, but 11.6% of the time they take more than 90 minutes and 2% of the time more than two hours. By then an attacker with 10% of the hashing power would have a 0.024% chance of reversing the payment.
The same payment sent in the halving block of April 20, 2024, when the median rate was 200 sat/vB, would have cost 28,200 satoshis, about $22.56 at the same assumed price.
The equivalent ether transfer is quicker at both ends. A plain transfer uses 21,000 gas; at the base fee of about 0.042 gwei seen on November 2, 2026, plus a 0.01 gwei tip, it costs 21,000 × 0.052 = 1,092 gwei, or about 0.0000011 ETH, a quarter of a cent at an assumed $2,300 per ether. In quiet conditions it is usually included within a slot or two, 12 to 24 seconds, and finalized roughly 13 minutes later. Live prices for both coins are on the bitcoin and ethereum pages.
What are the risks and limits of blockchains?
A blockchain secures the ledger, not the asset's price or your keys, and its permanence cuts both ways: mistakes and thefts are as irreversible as valid payments.
- No undo. A payment to a wrong address or an unsupported network cannot be charged back; sent crypto to the wrong network covers what can and cannot be recovered.
- Smaller chains are easier to attack. Security scales with the hashing power or stake behind a chain, so a small network can be outmined with rented hardware.
- Concentration. A few mining pools produce most Bitcoin blocks and a few large providers run a big share of Ethereum validators. None of them can change the rules, but they could delay or censor transactions for a time.
- Bugs. The 2013 split came from a database setting, and smart contracts on Ethereum and other chains have been exploited for large sums; what is DeFi covers that layer.
- Limited capacity. Base layers process a limited number of transactions, so fees jump when demand spikes; layer 2 networks move activity off the main chain at the cost of extra trust assumptions.
None of this makes the coins on a blockchain safe stores of value. Crypto is volatile and can lose most or all of its value even when the network underneath works exactly as designed, and the bitcoin halving guide shows how much bitcoin's price has fallen after past peaks.