Consensus: How Strangers Agree on Truth

How thousands of computers worldwide reach agreement without trusting each other — the Byzantine Generals Problem, proof-of-work, proof-of-stake, and what they buy you.

14 min · beginner · part of Blockchain: The Trust Machine

The Byzantine Generals Problem

Imagine you are a Byzantine general. You and other generals surround an enemy city. You must all attack together to win, or all retreat to survive. A mixed response means defeat. The problem: you can only communicate through messengers who might be intercepted, and some of the other generals might themselves be traitors trying to cause confusion. How do you reach a coordinated decision when you cannot trust the messengers or even all of your fellow generals? This is not just a colorful story. It is a real, formally defined problem in computer science, posed by Leslie Lamport, Robert Shostak, and Marshall Pease in their 1982 paper "The Byzantine Generals Problem." They proved that for a system with n participants where m of them might be malicious, no consensus is possible if m ≥ n/3. In other words, in any distributed system, more than two-thirds of participants must be honest for the system to function. For decades, the consensus mechanisms developed for distributed databases (Paxos, Raft) assumed that participants might fail but would not be actively malicious. They worked well in well-behaved environments like a single corporate network. But for an open, permissionless network where anyone could join — and where some participants would inevitably try to cheat — these were inadequate. Bitcoin's most famous breakthrough was solving the Byzantine Generals Problem at internet scale, in a permissionless network, in a way that was practical and economic. Whether you call this Nakamoto Consensus or Bitcoin Consensus, it is a major contribution to distributed computer science. The mechanism is proof-of-work mining, and the rest of this lesson digs into how it actually works and what the alternatives are.

Also in this lesson

Proof-of-Work in Depth
Proof-of-Stake: The Alternative
Other Consensus Mechanisms
For Deeper Reading

Key terms

Consensus mechanism: The protocol by which a distributed network of computers agrees on a single shared state. Bitcoin uses proof-of-work; Ethereum uses proof-of-stake.
Byzantine Generals Problem: A 1982 formal problem in distributed computing about reaching consensus when participants may be unreliable or malicious. Proven that consensus requires more than 2/3 honest participants.
Proof-of-work (PoW): A consensus mechanism where participants compete to solve computational puzzles, with the winner adding the next block. Used by Bitcoin. Energy-intensive but battle-tested.
Proof-of-stake (PoS): A consensus mechanism where validators lock up cryptocurrency as collateral to earn the right to validate blocks. Used by Ethereum since September 15, 2022. Reduces energy use ~99.95% vs PoW.
Nonce: A 32-bit number that miners change repeatedly while searching for a valid block hash. The nonce is included in the block header.
Difficulty: A measure of how hard it is to find a valid block hash. Adjusts every 2,016 blocks on Bitcoin (~2 weeks) to maintain ~10-minute block times. Currently ~144.4 trillion (Feb 2026).
Slashing: The PoS penalty mechanism where misbehaving validators have their staked cryptocurrency partially destroyed. Designed to make double-signing and other attacks economically unattractive.
Finality: The point at which a transaction is considered irreversible. Bitcoin offers probabilistic finality (~6 confirmations / 1 hour). Ethereum PoS offers explicit finality after ~12.8 minutes.
Validator: A participant in a PoS network who stakes cryptocurrency to earn the right to propose and attest to blocks. Ethereum requires 32 ETH to become a validator.
The Merge: Ethereum's transition from proof-of-work to proof-of-stake on September 15, 2022. One of the largest live software upgrades in computing history.
Delegated Proof-of-Stake (DPoS): A consensus mechanism where token holders elect a small set of validators (often 21-100). Used by EOS, Tron, Steem. Faster but more centralized than PoS.
Slot and epoch (Ethereum): Ethereum PoS divides time into 12-second slots, with 32 slots forming an epoch (~6.4 minutes). One validator is chosen per slot to propose a block.

Continue this lesson — 4 more sections in the CryptoBipto app.

Open lesson

Educational only — not financial advice.