03 — Consensus Mechanisms

How blockchains reach distributed agreement on transaction ordering. No consensus, no blockchain.

1. Taxonomy

2. Proof-of-Work (PoW)

Originated in Hashcash (Adam Back, 1997). Applied to blockchain by Satoshi (2008).

Mechanics

The block header has a nonce field. The miner varies the nonce until:

Target = 2^256 / difficulty. Difficulty adjusts every 2016 blocks (Bitcoin) to maintain ~10 min/block.

Heaviest chain (Nakamoto consensus)

A node chooses the chain with the most accumulated work (not length). Reorgs are possible up to probabilistic finality (6 confirmations ≈ ~99.99% safe).

Algorithms

Pros / Cons

Pros:

• Battle-tested (Bitcoin 15+ years).
• Sybil resistance via real-world cost (energy).
• Conceptually simple.
• Requires no social trust (just hashes).

Cons:

• Massive energy consumption (~150 TWh/year for Bitcoin globally, ~Argentina).
• Mining centralization (China + USA + Kazakhstan dominate; pools concentrate).
• Reorg risk for small chains.
• Slow finality (~60 min probabilistic for Bitcoin).

State in 2026

Bitcoin: ~700 EHs hashrate. Halving #4 in 2024 (3.125 BTCblock). Next halving 2028.

Ethereum: completed The Merge in Sep 2022; PoW deactivated. Ethereum Classic remains PoW.

3. Proof-of-Stake (PoS)

Stake (tokens) as collateral. Validators are selected proportionally to stake. Misbehaving = slash (lose stake).

Variants

Chain-based PoS (Nakamoto-style PoS)

A validator selected by an algorithm (VRF, randomness) constructs a block. Chain selection like PoW but weighted by stake.

Examples: Peercoin (2012, the first), Cardano (Ouroboros).

Committee-based / BFT-PoS

A subset of validators (committee) runs BFT consensus in each slot. Deterministic finality.

Examples: Cosmos (Tendermint), Algorand (BA*), Ethereum 2.0 (Casper FFG + LMD-GHOST).

DPoS (Delegated PoS)

Token holders vote for a small set of delegates (21 EOS, 100 Tron). Faster but lower decentralization.

Examples: EOS (legacy), Tron, BSC (partial), Steem.

Ethereum 2.0 (post-Merge)

• Validators: 32 ETH stake per validator slot. Currently ~1M active validators.
• Beacon Chain: PoS chain consensus.
• Slot: 12 seconds. Validator chosen via RANDAO to propose a block.
• Epoch: 32 slots (~6.4 min). Attestations finalize epochs.
• Casper FFG (Friendly Finality Gadget): 2 epochs without challenge → finality.
• LMD-GHOST: fork choice within the finalized chain.
• Slashing: surround vote / double vote → forfeit ETH (1-100% based on severity + correlation).

Slashing conditions

• Double signing: signing two blocks in the same slot.
• Surround vote: attestation conflicting with a prior one.
• Inactivity leak: < 2/3 online → progressive penalty for offline validators (forces finality recovery).

MEV under PoS

Proposer-Builder Separation (PBS): the validator (proposer) does not build the block; it receives a pre-built one from a builder that outbids in the mev-boost auction.

Flashbots mev-boost: software that separates builder from proposer. ~90% of Ethereum blocks use mev-boost. Dominant builders: Beaverbuild, rsync, Titan.

Inclusion lists (EIP-7547, under discussion): the proposer can force the inclusion of certain txs.

Pros / Cons

Pros:

• Energy ~99.9% lower than PoW.
• Faster finality (in BFT-PoS, seconds).
• Slashing punishes bad actors.
• Explicit economic security (\(1B+ staked = \)1B+ cost to attack).

Cons:

• Nothing-at-stake (theoretical; mitigated by slashing).
• Wealth concentration → power concentration.
• Initial distribution problem ("rich get richer").
• Long-range attack: a validator with old stake can forge history (requires weak subjectivity).
• Liquid staking centralizes (Lido controls 30%+ of ETH staking in 2026).

4. Tendermint / Cosmos

Jae Kwon, Ethan Buchman 2014–2016. Tendermint BFT + Cosmos SDK.

Tendermint consensus

Round-based BFT:

1. Propose: leader sends block.
2. Prevote: validators vote.
3. Precommit: validators commit if 2/3+ prevoted.
4. Commit: block finalized if 2/3+ precommitted.

Each round ~2-7s. Instant finality after commit.

Requires \(n \geq 3f + 1\) for \(f\) Byzantine.

Cosmos SDK

A Go framework for building app-specific blockchains. Adoption: Osmosis, Injective, Akash, Kava, Stargaze, Sei, Celestia.

IBC (Inter-Blockchain Communication)

Protocol for transfer + messaging between Cosmos chains. Standardized, native. See 11-bridges-interop.md.

5. HotStuff (Aptos, Sui, Diem legacy)

Maofan Yin et al. 2018 (Facebook Diem research). Refinement of PBFT with:

• Linear communication: \(O(n)\) msgs per round (vs \(O(n^2)\) PBFT).
• Responsiveness: progress at network speed, not timeout-driven.
• Chain quorum certificates (QCs): pipeline 3 rounds.

Adoption:

• Aptos (DiemBFT v4 = HotStuff variant).
• Sui (Narwhal + Bullshark — DAG + HotStuff).
• Monad (parallel EVM + HotStuff-style).
• Diem (discontinued 2022).

6. Algorand

Silvio Micali 2017 (Turing Award 2012). BFT-PoS with VRF.

Mechanics

1. Block proposer chosen via VRF (verifiable random function): each validator computes; only winners (top of probabilistic threshold) propose.
2. Committee also chosen via VRF.
3. BA★ (Byzantine Agreement star): VRF-based committee runs Byzantine agreement.

Properties

• Pure PoS: stake is just entry (no slashing in the original; added later).
• Finality: instant (committee approves block).
• Fork-free: VRF + BA★ guarantees a single chain.

Adoption

ASA (Algorand Standard Assets), Folks Finance, Algofi (defunct). USDC native on Algorand. Institutional adoption > retail.

7. Avalanche

Emin Gün Sirer et al. 2018. Family of consensus protocols.

Avalanche / Snowman

Repeated sub-sampled voting:

1. Validator queries random \(k\) peers.
2. Asks "what's the head?".
3. If \(\alpha\) of \(k\) agree → adopt.
4. Repeat \(\beta\) times → confident finalization.

Probabilistic finality in ~1s.

Three chains

• X-chain (Avalanche Native): UTXO + Avalanche consensus.
• C-chain (Contract): EVM-compatible, Snowman.
• P-chain (Platform): validator management + subnet coordination.

Subnets

Custom blockchain spawn (now "L1s" rebranding 2024). Permissioned or public.

8. Cardano

Charles Hoskinson (Ethereum cofounder), IOHK. Launch 2017.

Ouroboros

Proof-of-Stake research-led family of protocols (Praos, Genesis, Hydra).

• Slots: 1 second.
• Epoch: 432k slots (5 days).
• Slot leader: chosen via VRF, weight-proportional to stake.
• Coin age: not used (vs older PoS designs).

eUTXO model

Extended UTXO: outputs can carry datum + script (validator). Plutus = Haskell-based smart contract platform.

Stake pools

Delegate stake (ADA) to a pool. The pool operator runs a node. Rewards distributed proportionally. Saturation point (~70M ADA per pool) discourages concentration.

Historical pace

Academic research focus: papers published before code (IOHK + Univ Edinburgh + Lancaster + Tokyo). Branded "evidence-based" but criticized for slow shipping (smart contracts only in Sep 2021).

9. Polkadot

Gavin Wood (Ethereum cofounder). Launch 2020.

Architecture

• Relay Chain: PoS, BFT (GRANDPA finality + BABE block production).
• Parachains: independent chains plugged into the relay; share security.
• Auctions: leases for parachain slots (1-2 years).
• Crowdloans: locked DOT lent to parachains.

Crypto

• BABE: block production via VRF, Ed25519.
• GRANDPA: finality gadget. Asynchronous safety + dynamic availability.

Notable parachains

Moonbeam (EVM), Acala (DeFi), Astar, Phala, Kusama (canary), Centrifuge.

Polkadot 2.0 (2024+)

Shift from auctions to agile coretime — flexible compute purchase. JAM (Join-Accumulate Machine) — modular execution model.

10. Solana

Anatoly Yakovenko + Eric Williams 2017–2020.

Proof-of-History (PoH)

Not a consensus mechanism; it's pre-consensus ordering.

Sequential VDF: a validator runs \(H(H(H(\dots)))\) producing a chain of verifiable timestamps. Allows validators to agree on order without communication.

T0: hash0
T1: hash1 = SHA256(hash0)
T2: hash2 = SHA256(hash1 || tx_at_T2)
...

Tower BFT

PBFT variant using PoH timestamps. The slot leader rotates per ~400ms slot.

Pros / Cons

Pros:

• Throughput: ~3000-5000 TPS sustained, peaks 65k.
• Latency: ~400ms block, ~1.5s finality (32 slots TLF).
• Low fees: ~US$0.0001/tx.
• Parallel execution (Sealevel).

Cons:

• Validator HW expensive (high decentralization cost; ~3000 validators).
• Stability: historical outages (Sep 2021 16h, Jan 2022, Jun 2022, Feb 2023, etc.).
• Single client (Solana Labs) — Firedancer (Jump Trading) client in testnet 2024, mainnet 2025.

Sealevel

Parallel execution: txs declare read/write accounts; conflicts detected; non-conflicting ones parallelized.

11. DAGs

Directed Acyclic Graph instead of a linear chain.

Hashgraph (Hedera)

Mance Harmon + Leemon Baird 2016. Gossip about gossip + virtual voting. Patented (closed-source until 2018, open since then). Permissioned (Hedera Council).

IOTA Tangle

A tx pre-attaches 2 other txs. PoW per tx. Coordinator required for security until 2.0. Adoption: mostly IoT pilots.

Conflux

Tree-Graph hybrid. Inclusive blocks (uncle-style but weighted properly).

Avalanche Snowman

Strictly speaking a chain not a DAG, but Avalanche X-chain (DAG-based).

Aleph Zero

DAG + AlephBFT. Sub-second finality. Privacy via ZK on roadmap.

12. Proof-of-Space / PoSt

Chia

Bram Cohen (BitTorrent inventor) 2017.

• Plotting: precompute large files (proofs of space) on disk.
• Farming: scan plots for valid proofs at each challenge.
• VDF: Wesolowski VDF for timing between blocks (Time Lord).

Tagline: "green Bitcoin" (but plotting causes lots of SSD writes).

Filecoin

Protocol Labs (Juan Benet) 2020.

• Proof-of-Replication (PoRep): prove unique storage of a file.
• Proof-of-Spacetime (PoSt): prove continuous storage over time.
• Expected Consensus (EC): leader election weighted by storage power + VRF.

Spacemesh

PoSt + PoET (Proof-of-Elapsed-Time, Intel SGX-based originally, later software). Smesher = home miner. Launched 2023.

13. Proof-of-Authority

Validators are named, trusted entities. No mining, no significant stake.

Examples

• POA Network (legacy).
• VeChain Authority Masternodes.
• xDai (Gnosis Chain) PoA → POSDAO.
• BNB Chain (BSC): 21 validators initially Binance-selected; now some elections.
• Polygon PoS: 100 validators, primary Polygon Labs.

Pros / Cons

Pros: Fast, cheap, predictable. Cons: Centralized — a regulatory attacker just needs to pressure the validators.

Common pattern: launch PoA → migrate to PoS or hybrid over time.

14. Hybrid and exotics

Decred

PoW (Blake-256) + PoS (stakeholders vote on PoW blocks). Treasury governance built-in.

Filecoin

PoSt + EC (above).

Helium

Proof-of-Coverage. Miners demonstrate wireless coverage. Became a Solana token in 2023.

Internet Computer (ICP)

Dfinity. Threshold signatures + chain-key cryptography. Subnet-based scalability. Reverse-gas (developer pays).

Ethereum Casper FFG + LMD-GHOST

Casper FFG = finality gadget; LMD-GHOST = fork choice; Casper-FFG-on-top finalizes.

15. Finality

16. Practical comparison

(Stats ~May 2026, subject to change.)

17. Selecting a consensus for a use case

18. Cross-reference

• Detailed chains: 04-l1-bitcoin.md, 05-l1-ethereum.md, 06-l1-alt.md.
• L2 / rollups (built atop): 07-l2-scaling.md.
• Cryptographic primitives (VRF, VDF, signatures): ../cryptography/.
• MEV: 12-tokenomics.md §MEV.
• Slashing incidents: 14-incidents.md.