Learn · 01

What is a blockchain,
really?

Strip away the tokens and the hype, and a blockchain is a very specific engineering idea: a list of records that is expensive to rewrite, cheap to verify, and does not require anyone's permission to read. This page explains what that means, what makes it work, and — just as importantly — what it cannot do.

What it is

A blockchain is an append-only list of records, called blocks, where each block commits to the one before it by including the previous block's hash in its own header. If anyone changes an earlier block, its hash changes, which breaks the hash the next block expects, which breaks the one after that. The whole chain "falls over" if you try to tamper with it, and any reader can detect the tampering by recomputing the hashes.

That is the entire cryptographic idea. Everything else — mining, consensus, tokens, smart contracts, DeFi — is built on top of this one property.

Plain-language version

A blockchain is a notebook where every page contains a fingerprint of the previous page. If you rip out page 5 and replace it, page 6 stops matching it, and everyone notices.

The three ingredients

Every blockchain, from Bitcoin to a private enterprise ledger, combines the same three mechanisms:

  1. Hashing. A function that turns any data into a fixed-length fingerprint. Deterministic, one-way, collision-resistant.
  2. Digital signatures. A way for one party to prove that they authorised a specific record, without revealing a secret.
  3. Consensus. A rule for deciding which version of the chain is the real one when two versions exist.

If any of these is missing, the system is not a blockchain — it is just a database with extra steps.

1 · Hashing

A cryptographic hash function takes any input and returns a fixed-length output. SHA-256, the most common, returns 256 bits — 64 hexadecimal characters. Three properties make it useful:

  • Deterministic. The same input always gives the same output.
  • One-way. Given an output, you cannot recover the input faster than by brute force.
  • Collision-resistant. Finding two different inputs with the same output is computationally infeasible.

In a blockchain, hashes are used in two places. First, as a content-address: the hash of a file is a compact way to commit to its exact bytes. Second, as a link: each block's header contains the previous block's hash, chaining the history together.

# The hash of "hello" in SHA-256
2cf24dba5fb0a30e26e83b2ac5b9e29e1b161e5c1fa7425e73043362938b9824

# The hash of "hello." — one extra character
ce06092fc078ba9dd364a3bb66a1bf35a87b9db2c6c1e22b63bea8d8d18e4c5e

# Completely different. That is the point.

Merkle trees

When a block contains many transactions, hashing each transaction and then hashing pairs of hashes upward produces a single value called the Merkle root. This root goes into the block header, so a single hash commits to every transaction in the block. A verifier can prove that a specific transaction is in the block by showing a logarithmic-sized path of sibling hashes, rather than downloading the whole block.

2 · Digital signatures

A signature scheme has three parts: a key generation algorithm that produces a public/private keypair, a signing algorithm that uses the private key, and a verification algorithm that uses the public key. The security property is simple: given a public key and a message and a signature, anyone can check that the holder of the matching private key signed that exact message. Nobody else can produce such a signature.

In a blockchain, signatures are what authorise transactions. When you send Mera, you sign a message describing the transfer with your private key. Every other node can verify the signature against your public key without ever seeing your private key. That is what makes the system trustless: nobody has to trust you, they only have to check a mathematical fact.

Keys are not passwords

A lost password can be reset. A lost private key cannot. There is no "forgot my key" mechanism in a blockchain, because there is no authority to reset it. In MLabChain, the wallet file is encrypted with a password, but if you lose both the file and the password, the balance is gone.

3 · Consensus

The hardest problem. If two miners produce a valid block at the same height, the network has to decide which one is the "real" block. Different chains answer this differently.

MechanismHow it decidesTrade-off
Nakamoto PoW Longest chain by cumulative hash work wins. Simple, battle-tested, energy-heavy.
BFT / PBFT A supermajority of validators vote on each block. Fast finality, requires a known validator set.
Proof of Stake Validators are chosen in proportion to their stake. Energy-light, but "stake" is a new centralisation axis.
PoA / committee A named set of authorities signs blocks. Fast, cheap, but trust is explicit.

MLabChain uses a hybrid: cumulative-work fork choice for block sealing (Nakamoto-style PoW, at low difficulty), plus an Ed25519-signed block proposal that pins who produced each block. Scientific rewards are a separate mechanism altogether — see Proof-of-Scientific-Work.

What a blockchain is not

The word has been stretched to cover a great deal that it does not mean. It is worth being specific.

A blockchain does

  • Provide a tamper-evident history.
  • Allow anyone to verify a record without trusting the author.
  • Let multiple parties agree on a shared state without a central server.
  • Survive the loss of any individual node, if the network is large enough.

A blockchain does not

  • Store large data efficiently. It is an index, not a warehouse.
  • Make bad data good. Garbage in, immutable garbage out.
  • Solve governance. Someone still has to decide what counts.
  • Guarantee privacy. Public chains are public.
  • Remove trust. It relocates trust to different places.

Variants of "blockchain"

Three broad families, in increasing order of decentralisation:

  • Private / enterprise. A consortium runs the nodes. Fast, cheap, permissioned. Trust is in the consortium. Hyperledger Fabric is the canonical example.
  • Public permissioned. Anyone can read, a known set can write. Good for regulated use cases.
  • Public permissionless. Anyone can read and write, subject to economic rules. Bitcoin, Ethereum, and most "crypto" chains.

MLabChain is in the first family. It is a local, single-writer ledger with an optional multi-node testnet mode. It is not permissionless and does not pretend to be.

Why this matters for MLabChain

The value of a blockchain for scientific work is not decentralisation for its own sake. It is that a researcher can prove what they did, without asking anyone's permission and without depending on a service that might disappear.

When you register an ML training run on MLabChain, you get a signed, hash-linked record. Six months later, you can re-run the training from the recorded config and check that the metrics still match. Nobody else has to be involved. No server has to be online. That is a modest, useful property, and it is the reason the project exists.

Take-away

A blockchain is a hash chain plus signatures plus a consensus rule. It is a specific technical answer to a specific technical problem. The rest is application.