[Beginner] — the vocabulary, the one technical claim underneath all of it, and how to tell the engineering from the marketing. Assumes nothing.
The kid version first
Normally, when you own something digital — money in a bank, a username, an item in a game — one company keeps the list that says you own it. They can edit that list. You trust them not to.
A blockchain is a way for thousands of strangers to each keep a copy of the same list, and agree on what it says, without any of them being in charge. That’s it. That’s the whole invention.
Everything else — coins, NFTs, DeFi, DAOs — is an application built on “we can all agree on a list without a boss.”
The terms, disentangled
The words get used interchangeably by people selling things. They are not the same:
| Term | What it actually means |
|---|---|
| Blockchain | A specific data structure: a chain of blocks, each committing to the previous by hash. Not inherently decentralised — a company can run one alone (and usually shouldn’t) |
| Distributed ledger | The general category. A shared, replicated database. Blockchain is one design |
| Cryptocurrency | A blockchain whose ledger tracks token balances, with a native asset used to pay for writes |
| Web3 | A marketing-adjacent umbrella term for applications built on public blockchains. Coined by contrast with “Web2” (platform-owned) |
| Smart contract | A program deployed to a blockchain, whose execution every node reproduces and agrees on |
| DeFi | Financial applications (lending, exchange, derivatives) as smart contracts |
| NFT | A token whose units are individually distinct rather than interchangeable |
| DApp | An application whose backend is partly a smart contract |
“Web3” is the weakest of these terms — it describes a vibe and an investment thesis more than a technology. The engineering underneath is real; the term is not load-bearing. This folder uses it because that’s what the field is called, not because it’s a good name.
The one technical claim
Strip away everything else and a blockchain makes exactly one novel offer:
Verifiable state, agreed by parties who don’t trust each other, with no administrator.
Three parts, all necessary:
- Verifiable — anyone can independently check the current state is the correct consequence of the history. You don’t take anyone’s word for it
- Agreed — everyone converges on the same answer, even under network delays and active attackers
- No administrator — nobody can unilaterally rewrite it, censor it, or switch it off
Remove the third and you don’t need a blockchain. A Postgres database with an audit log and signed writes gives you (1) and (2) at a millionth of the cost. This is the single most useful test to apply to any blockchain proposal, and it kills most of them → what blockchains are actually good for.
What you give up to get it
The trade is real and steep:
- Throughput. Ethereum settles on the order of 15–100 transactions per second at L1. Visa handles thousands. Every node executes every transaction — that’s the point, and it’s also the ceiling
- Latency. Finality is seconds to minutes, not microseconds
- Cost. You pay per unit of computation and per byte of storage, forever, to everyone
- Privacy. Public chains are public. Your balance and every transaction you have ever made are readable by anyone → zero-knowledge proofs are partly an answer to this
- Irreversibility. No chargebacks, no password reset, no support line. Lose the key, lose the asset. This is a feature and a catastrophe depending on the day
A blockchain is a very slow, very expensive, very public computer that nobody owns. You pay all of that for the “nobody owns” part. When that property is worth it, nothing else does the job. When it isn’t, you have built a bad database.
Public, private, permissioned
- Public / permissionless (Bitcoin, Ethereum) — anyone can read, write, and run a node. The decentralisation claim is meaningful
- Permissioned / consortium (Hyperledger Fabric, R3 Corda) — a known set of organisations run the nodes. Useful in narrow inter-company settlement cases; mostly a replicated database with extra steps, and the enterprise-blockchain wave of 2016–2020 largely ended in that realisation
- Private — one company, one chain. Almost always the wrong tool
This folder is about public chains, because that’s where the interesting engineering and the actual use cases are.
The layers, so you know where you are
Application DeFi protocols, NFT markets, DAOs, games
─────────────────────────────────────────
Contract Smart contracts (Solidity, Rust, Move)
─────────────────────────────────────────
Execution The VM — EVM, SVM, WASM. Deterministic, metered
─────────────────────────────────────────
Consensus Who gets to append the next block, and when is it final
─────────────────────────────────────────
Network P2P gossip, mempool, node discovery
─────────────────────────────────────────
Data Blocks, hashes, Merkle trees, signatures
The rest of this course walks up that stack: 01-foundations covers data through consensus, 02 covers execution, 03 the contract layer, 07 the top.
Key insight
The blockchain is not a better database. It’s a worse database with one property no database has — that it keeps working when the parties running it actively want to cheat each other. Every design decision in this folder, and every cost, follows from buying that one property.
Related
- the double-spend problem — why this needed inventing at all
- the honest assessment — read this early, not last
- what makes distributed systems hard — the field this is a subfield of
- cryptography — the primitives this is assembled from
Source: [reference] — Aug 2026.
Software engineering
Infrastructure & security
Computer science
Hardware & engineering