A field guide to blockchain consensus for people who own crypto but were never told how these networks actually decide what is true. Five networks, five very different answers, and what each one is really good for.
Before we can talk about how five networks reach agreement, we need to be honest about the question they are all answering. It is a surprisingly old one: how do you move value, or agree on a record, when nobody at the table fully trusts anybody else?
Think about what happens when you send someone money today. You tap a button, and somewhere in the background a bank, a card network, or a payment app such as PayPal updates a list. That list says you have a little less and they have a little more. The whole system works because everyone agrees to treat that institution's list as the official one. The bank is the middleman, the trusted party in the middle that keeps the master record and settles disputes.
That arrangement is convenient, and most of the time it is fine. But it comes with quiet costs. The middleman can freeze your account, reverse a payment, make an error, charge a fee, close on a holiday, or simply tell you no. You are trusting one organization to be honest, accurate, and available. For a lot of the world, and a lot of situations, that trust is not a safe assumption.
So why not skip the bank and just send digital money directly, the way you send a photo? Here is the catch, and it is the heart of the whole field. A photo is just data, and data copies perfectly. If money were just a file, you could send the same file to ten people at once, or send it to a friend and still keep a copy for yourself. Spending the same unit of money twice is called the double-spend problem, and it is the reason digital cash did not work for decades.
In the bank world, double spending is impossible for a boring reason: there is one master ledger, and the bank checks it before every payment. A ledger is simply the running list of who owns what. One authoritative list, one source of truth, problem solved. The cost, again, is that you have to trust whoever holds the list.
If you remove the trusted middleman, who keeps the ledger? If everyone keeps their own copy, whose copy is correct when two copies disagree? Answer that, and you have invented a blockchain.
A blockchain is one shared ledger that thousands of strangers maintain together, with a strict set of rules that makes them converge on a single agreed version, batch by batch. Each batch of transactions is a block, and each block is mathematically chained to the one before it, which is where the name comes from. Tampering with an old block would break every block after it, in plain view of everyone. There is no master copy to attack, because everyone has the same copy.
It is tempting to say blockchains are "trustless." That is not quite right, and the real story is more interesting.
Blockchain does not eliminate trust. It redistributes it.
Instead of trusting one bank not to make a mistake, freeze your funds, or quietly change the record, you are now trusting something different: that a majority of the network's validators will not coordinate to cheat at the same time. A validator is just a participant that helps check transactions and agree on the official record. Trust did not disappear. It moved, from a single named institution you can sue, to a large crowd of independent participants who would all have to collude to break the rules.
That shift is the entire point, and it is also the entire debate. How many validators do you need? How do you stop them colluding? How do you make cheating expensive and honesty profitable? How fast can a crowd of strangers agree without a boss? Every network in this guide is a different answer to those questions. The procedure each one uses to get that crowd to agree has a name: consensus.
Consensus is the mechanism that solves the problem from Section 01. It is the agreed procedure a leaderless network uses to settle, over and over, on one official version of the ledger that everyone accepts.
Picture a few thousand strangers around the world, each holding a copy of the same notebook, with no manager and no head office. Every few seconds a fresh batch of transactions needs to be written into every copy, in the same order, with no disagreement. There is no one to break a tie. The rules themselves have to produce the tie-break. That set of rules is the consensus mechanism.
A good consensus mechanism has to do two hard things at once. It has to let honest participants agree quickly even though messages arrive out of order and some participants are offline. And it has to make sure a dishonest minority, or even a fairly large dishonest group, cannot force a false version of the record onto everyone else. Doing both, among strangers, with real money on the line, is genuinely difficult. That difficulty is why so many different designs exist.
Every consensus design is balancing four things that pull against each other: security (how hard it is to attack), speed (how fast payments settle), decentralization (how many independent participants, and how easy it is to join), and programmability (whether the network can run apps, not just move a coin). You cannot turn all four to maximum at once. Push one up and you usually give a little back somewhere else.
Three of those four levers pull against each other so reliably that the tension has a nickname: the blockchain trilemma. Those three are security, decentralization, and scalability (the speed-and-volume one). The rough rule is that you can have any two in full, but maxing all three at the same time is the hard, still-unsolved part. Programmability sits apart, a separate design choice rather than a third corner of the triangle.
This is why there is no single "best" network, only different bets. The five networks ahead were chosen precisely because they make those bets differently, not because of how large they happen to be at any moment. Read them as five honest attempts at the same impossible-to-perfect problem.
"Final" does not mean the same thing on every network. Four flavors turn up in the pages ahead, and it is worth keeping them straight.
Spend real electricity to make cheating expensive. Security and openness come first, and speed comes last.
*Decentralization is a theoretical ceiling. In practice a few mining pools command most of the hashrate, which concentrates who actually builds blocks.
The original design. Security and openness first, speed last, paid for with electricity.
Imagine a worldwide lottery that runs every ten minutes, except you cannot just buy a ticket with money. The only way to enter is to spend electricity. The players are called miners, computers competing for the right to add the next block. They all race to solve the same puzzle, and the puzzle has no clever shortcut. You simply guess, trillions of times per second, until someone stumbles onto a winning answer.
The puzzle is this: find a number, called a nonce, that makes the block's digital fingerprint, its hash, fall below a target value. There is no way to calculate the answer in advance. You can only keep guessing, which means the only way to win is to do enormous amounts of work. That work is called Proof of Work, and the wasted-looking electricity is the entire security model. Rewriting history would mean redoing all that work, faster than every honest miner on Earth combined. That is what makes cheating not just hard but wildly expensive.
Finality is the moment a payment can no longer be undone. Bitcoin's version is unusual: a transaction is never stamped "certified final." Instead it becomes harder and harder to reverse as more blocks pile on top of it. One block deep, it is very likely permanent. Six blocks deep, about an hour later, reversing it would be so costly it is treated as settled. This is why exchanges make you wait for confirmations on large amounts. The payment is not getting more "approved," it is getting more buried.
Each winning miner collects a block reward, newly issued bitcoin, plus the fees attached to the transactions they include. Playing by the rules pays a steady income. Attacking the network would cost more in electricity than it could ever steal, and would destroy the value of the very coins the attacker holds. The incentives are arranged so that honesty is simply the more profitable choice. That is the quiet genius of the design.
A censorship-resistant store of value and final settlement of large amounts. The thing it does best is simply being extremely hard to stop, change, or seize.
Fast everyday retail payments or rich on-chain applications. Bitcoin chose to be simple and unbreakable rather than fast and clever.
Replace the electricity with a cash bond, then turn the secured ledger into a programmable world computer.
*Decentralization is a theoretical ceiling. A large and growing share of staked ETH sits with a few staking pools and providers, which concentrates influence.
A world computer. Trades a little speed and a lot of simplicity for the ability to run programs.
Ethereum asks a sharp question: do we really need to burn electricity to make cheating expensive? Its answer is no, just make people put money on the line instead. Picture a voting pool where, to get a vote, you first post a sizable cash bond. If you vote honestly, you earn a little. If you are caught cheating, the pool keeps part of your bond.
To become a validator on Ethereum, you lock up 32 ETH as a security deposit.2 Locking up coins this way is called staking. The network randomly picks one validator to propose the next block, and a committee of others to check it and vote that it is valid. Voting yes on a good block is called an attestation. Behave well and you collect modest rewards. Try to cheat by signing two conflicting versions of events, and the network destroys part of your stake. That punishment is called slashing, and it replaces electricity as the cost of dishonesty. Going offline is treated far more gently: you miss rewards and slowly leak a little stake while away. Slashing is reserved for provable foul play. Your own money is the thing at risk.
This is the distinction that trips people up. A new block lands about every twelve seconds, so your transaction usually appears on the chain quickly. But appearing is not the same as being final. Ethereum reaches full finality in two steps. It marks a checkpoint about every six and a half minutes, and a block is only locked in for good after two checkpoints have passed. Two of those, back to back, is where the familiar thirteen-minute figure comes from.3 Fast to show up, slower to become truly permanent. Both numbers are normal, they simply describe different moments.
Ethereum's real breakthrough is not the consensus method, it is what the consensus secures. Ethereum runs smart contracts, self-executing programs that live on the chain and run exactly as written, with no company in the middle able to alter or stop them. A smart contract can hold funds, enforce an agreement, or run an entire application. This is why Ethereum is described less as digital gold and more as a world computer. Most of what people mean by "crypto apps," lending, stablecoins, digital collectibles, exchanges that run without a company, was built on this foundation.
Smart contracts and an entire ecosystem of applications built on top of them. If a crypto product is doing something more complex than sending a coin, it very likely started here.
Maximum raw throughput at rock-bottom cost on the base layer. That job is increasingly handed to layer 2 networks, extra layers that bundle activity and settle back to Ethereum.
Let a circle of trusted validators agree in seconds, settling value with final, low-cost simplicity.
A circle of trusted validators that settles value in seconds. No mining, no race.
XRP does not use a race at all. Its method belongs to a family called Byzantine Fault Tolerance, the ability of a system to reach a correct, shared answer even if some participants are broken or actively lying. The name comes from a classic puzzle about generals trying to coordinate an attack while some of them are traitors sending false messages. The specific recipe XRP uses is the Ripple Protocol Consensus Algorithm, or RPCA.
Picture a room of people, each of whom trusts a known list of respected colleagues. Every few seconds, they each propose which transactions should go in the next batch. They look at what everyone else proposed, drop anything that is not widely shared, and propose again. They keep adjusting through several quick rounds, and the bar for agreement rises each round, until at least 80% of their trusted colleagues have landed on the exact same set. At that point the batch is locked into the ledger. No puzzles, no electricity, just rapid rounds of "here is my list, let me see yours, let me adjust."
Here is the clever part that keeps it from breaking. Each validator follows a list of validators it chooses to trust, called a Unique Node List, or UNL. The safety of the whole system rests on a simple fact: these lists heavily overlap. Because almost everyone's trusted set shares most of the same members, the network does not split into two groups that each believe they reached agreement on different answers, as long as those trusted lists overlap enough. That overlap is the glue. A ledger splitting into two conflicting versions is called a fork, and keeping the trusted lists heavily overlapped is what prevents one. The safety is strong but conditional: it depends on that overlap holding, not on a law of physics. To make this easy to set up, the XRPL Foundation publishes a recommended validator list, so operators do not each have to assemble trust from scratch.
There is no mining reward and no staking payout here. All XRP was created at the start, so validators are not paid in new coins. They run for reliability and reputation, not a prize, and a tiny fee on each transaction is destroyed rather than paid out. On programmability, the picture is changing. The base ledger is gaining native programmability built on WebAssembly (a fast, portable way to run small programs), rolling out in deliberate phases: simpler programmable features are already live, while broader native smart-contract support is in active development on test networks. A separate EVM sidechain, a companion network that speaks Ethereum's language, also exists for Solidity-style apps. Even so, the honest framing holds: the base ledger still favours a focused set of fast, purpose-built primitives over being a general world computer like Ethereum, and full native programmability is maturing rather than finished.
Fast, cheap, final settlement of value, with cross-border payments as the headline use case. It is purpose-built to move money quickly.
Heavy, general-purpose app development on the base ledger, at least for now. Native programmability is arriving in phases rather than being a finished, all-purpose platform like Ethereum.
Timestamp every transaction the moment it arrives, so ordering is solved before anyone votes and the speed follows.
*Scalability is a theoretical peak. Multiple network outages have interrupted finality and reduced real-world throughput reliability.
Built for raw speed. Timestamps everything up front so validators never argue about order.
In most networks, the slowest and most argument-prone part of consensus is agreeing on the order of events. Whose transaction came first? Solana's insight is to settle the order before anyone votes. Picture a notary who stamps an exact time on every envelope the instant it arrives, before any of it goes into the pile. Later, nobody has to argue about sequence, because the stamps already prove it.
That notary is called Proof of History. It is not a consensus mechanism on its own, it is a clock. It works by running a continuous chain of computed timestamps, each one provably built on the last, so the network can show that event B happened after event A without validators ever comparing their watches. On top of that clock, Solana runs a Proof of Stake voting system of its own, called Tower BFT: validators stake coins and vote to confirm the already-ordered stream, with more stake carrying more weight. Going offline or missing votes costs a validator rewards. The clock does the ordering, the staked voting secures it.
Solana separates two moments, much as Ethereum does. A transaction is confirmed in well under a second, as soon as a supermajority of validators has voted for the block it sits in. That fast confirmation is what makes the network feel instant. But confirmation is not full finality. A block is only treated as finalized, rooted so deeply that the network will not roll it back, once about another thirty blocks have been built and voted on top of it, which takes roughly thirteen seconds.5 Sub-second is the confirmation you see; the firmer guarantee lands a little after. Both numbers are normal, they simply describe different moments.
The payoff is dramatic. Solana can process very high volumes of transactions, confirming in well under a second to a few seconds, at very low fees. For applications that need to feel as instant as a normal app, that matters.
Now the honest part. That speed has costs that are easy to skip past. Keeping up with the firehose of timestamped transactions demands powerful, expensive hardware, which means there are fewer validators than on Bitcoin or Ethereum. Fewer validators is a real decentralization trade-off. And the network is younger and has had outages, periods where it stopped producing blocks and had to be restarted by its operators.6 It is fast and steadily improving, but it does not yet have the long "nothing has ever stopped it" track record that Bitcoin can point to.
High-volume, low-cost applications at scale: fast trading, payments, and consumer-facing apps where speed and tiny fees are the whole experience.
The maximal "it has never once stopped" reliability story. Solana is betting that performance, refined over time, is worth a younger and more demanding design.
Drop the single chain of blocks. Let a gossiping web converge on order, fast and mathematically final.
*Both bars are theoretical ceilings. The asynchronous BFT guarantee is the strongest category on paper, but in practice it rests on a small, permissioned set of council-run nodes, not an open crowd, and real-world throughput has run well below the headline figure.
A fundamentally different shape. No chain of blocks at all, but a gossiping web.
Every network so far builds a single line of blocks, one at a time. Hedera throws that picture out. Its structure is a Directed Acyclic Graph, or DAG, a web of connected events that all flow forward in time and never loop back on themselves. Instead of one straight chain, imagine a spreading web.
Here is how it fills in. There is no winner and no leader. Instead, every node constantly gossips. When two nodes connect, each tells the other everything it has heard and exactly when it heard it. Crucially, they also pass along the fact that they passed it along, a record of who told whom, when. This is called gossip about gossip. Very quickly, every node ends up holding the same complete web of who-heard-what-and-when. And because they all share that identical history, each node can calculate how every other node would vote on the ordering of events, without sending a single additional message. Working out the result from shared history alone is called virtual voting, and skipping the extra round of messages is a big part of why it is fast.
This is the key divergence worth holding onto. A blockchain is sequential: one block at a time, built by one participant at a time, with everyone else waiting. A hashgraph is parallel: messages from everyone flow at once and get woven into a single agreed order afterward. The shape is a web, not a chain. On security, Hedera uses what is, on paper, the strongest category of Byzantine Fault Tolerance, called asynchronous BFT, and its finality is deterministic, settled in a few seconds with a mathematical guarantee rather than a probability that grows over time.
Speed and efficiency come with a deliberate choice about who runs the network. Hedera's nodes are operated by the Hedera Governing Council, a fixed group of up to several dozen large, named organizations that take turns running the infrastructure. This makes the network stable, accountable, and predictable, you know exactly who the operators are. But it is permissioned: you cannot simply spin up a council node the way anyone, anywhere, can start mining Bitcoin. That is the honest trade-off. Hedera deliberately limits how open its governance is, in exchange for speed, efficiency, and predictable performance.
Enterprise-grade, high-volume uses, payments, asset tokenization, and tamper-proof logging, where predictable performance and known, accountable operators are a feature rather than a flaw.
The permissionless, anyone-can-join-and-validate ideal that Bitcoin holds as its core value. Hedera trades that openness for performance and accountability.
This scorecard is shorthand, and it only works because the pages before it did the real explaining. Read each cell as a pointer back to a nuance you now understand, not as a ranking. None of these is "winning." They are built for different jobs.
| BitcoinBTC | EthereumETH | XRPXRP | SolanaSOL | HederaHBAR | |
|---|---|---|---|---|---|
| Consensus type | Proof of Work | Proof of Stake | Byzantine Fault Tolerance (RPCA) | Proof of History on Proof of Stake | Hashgraph (gossip + virtual voting) |
| Problem it solves best | Unstoppable, censorship-resistant store of value | Programmable money and on-chain apps | Fast, final settlement of value across borders | High-volume, low-cost apps that feel instant | Predictable high-volume use with named operators |
| Finality speed | ~1 hour (6 blocks, probabilistic) | ~13 minutes (economic finality) | ~3 to 5 seconds (final) | ~1 sec to confirm, ~13 sec to finalize* | ~3 seconds, often faster (deterministic) |
| Energy use | Very high | Low | Very low | Low to moderate | Very low |
| Decentralization (openness to join) | ** Most open; anyone can mine, though pools concentrate power | *** Broad validator set; staking can concentrate | Curated trusted-validator lists | Fewer validators due to hardware demands | Permissioned council; most curated |
The meter shows how open each network is to anyone joining as a validator: more bars, more open. It is a rough cue, not a grade. Every position here is a deliberate trade made to be better at something else.
Scorecard accurate as of June 2026. Ratings reflect theoretical capabilities balanced against observed network conditions. See asterisks for specific caveats.
There is no best consensus mechanism, only different bets on security, speed, decentralization, and programmability. When you understand which bet a network made, you understand what it is genuinely useful for, and you stop comparing a vault to a sports car as if one is simply wrong.
How these ratings are set. The bars and meters throughout this guide are our qualitative reading of recognized indicators, not a single published score, and no agreed cross-network scale exists. Decentralization draws on the Nakamoto Coefficient and stake or hashrate concentration (see the Edinburgh Decentralisation Index and clientdiversity.org). Security draws on cost-to-attack estimates and length of track record (see crypto51.app). Scalability draws on observed versus theoretical throughput and finality (see chainspect.app). These indicators move over time, so read the bars as direction, not measurement.
Each network's timing and mechanism claims were checked against primary or official sources. These figures are perishable, since networks change their parameters, so they were verified as of June 2026. Confirm current status before relying on a number.
As of June 2026. Confirm current status; consensus parameters and network conditions change.
Plain-English crypto education, so you understand what you own and know what to ask next.
Download the whole report as a printable PDF. Same content, formatted for print and offline reading.
Download the PDF