[Beginner] — the specific problem blockchains were invented to solve, why it’s hard, and why every prior attempt failed. Assumes note 01.
The kid version first
If I send you a photo, I still have the photo. Copying digital things is free — that’s the whole point of digital.
That’s a disaster for money. If I send you a digital coin and I still have it, I can send the same coin to a hundred people. The coin is worthless.
Banks solve this by keeping the only list that counts. You can’t spend twice because the bank checks. The problem blockchains solve is: how do you stop double-spending when there’s no bank?
Why it’s genuinely hard
The naive fix is obvious — everyone keeps a copy of the ledger, and rejects a second spend of the same coin. It fails immediately:
Order is the problem, not detection.
I broadcast two conflicting transactions at the same instant — coin X to Alice, coin X to Bob — to different parts of the network. Half the nodes see “X→Alice” first and reject “X→Bob”. The other half see the reverse. Both halves behave perfectly correctly and now permanently disagree about who owns X.
Detecting a double-spend is trivial. Agreeing on which one came first, globally, with no clock anyone trusts and no authority to ask — that’s the hard part. This is the same problem distributed consensus has always been, with one brutal extra constraint.
The extra constraint: open membership
Classical consensus (Paxos, Raft) assumes you know who the participants are. Five nodes, majority of five. That assumption does a lot of work — it lets you count votes.
A public blockchain can’t have it. Anyone can join. And the moment votes are counted per-identity in an open system, you get the Sybil attack: I create ten million identities and outvote everyone. Identities are free; that’s what “open” means.
The real problem is not double-spending. It’s Sybil-resistant consensus among an unknown, unbounded, mutually distrusting set of participants.
Everything expensive about blockchains — proof-of-work burning electricity, proof-of-stake locking up capital — exists to solve this one thing. You cannot count identities, so you count something that costs money.
What was tried before, and why it didn’t work
Digital cash is much older than Bitcoin. The failures are instructive:
| Attempt | Idea | Why it failed |
|---|---|---|
| DigiCash (Chaum, 1989) | Blind signatures for untraceable e-cash — cryptographically excellent | Needed a central issuer to prevent double-spend. The company went bankrupt in 1998, and the money went with it |
| B-money (Wei Dai, 1998) | Proposed a broadcast ledger with computational work | Never specified how nodes agree on the ledger. Left the hard part as an exercise |
| Bit Gold (Szabo, 1998) | Chained proof-of-work timestamps | Same gap — no mechanism to converge on one chain |
| Hashcash (Back, 1997) | Proof-of-work as email anti-spam | Not money, but supplied the mechanism Bitcoin borrowed |
Every pre-2008 design either kept a trusted issuer or hand-waved consensus. The pieces — hash chains, Merkle trees, digital signatures, proof-of-work — all existed by 1998. None of them was new in 2008.
What Bitcoin actually contributed
Satoshi’s 2008 paper assembled known parts into one mechanism that closed the gap:
- Make writing expensive — appending a block requires a proof-of-work solution, so influence is proportional to computation, not to identity count. Sybil attacks stop being free
- Make history cumulative — each block commits to the previous one, so rewriting block N means redoing the work for N and everything after it, while the honest chain keeps extending
- Give an unambiguous tiebreak — “the valid chain with the most accumulated work wins.” Nodes need no coordination, no voting round, no knowledge of the participant set. They apply a local rule and converge
That third point is the quiet genius. Consensus becomes a rule each node applies alone, rather than a protocol nodes run together. The order of transactions is decided by whoever wins the work lottery, and the network agrees on that ordering because the tiebreak rule is deterministic and self-evident from data everyone has.
The cost of the answer
The solution is real, and it is not free:
- Probabilistic finality. Under proof-of-work, “settled” means “expensive to reverse,” not “impossible.” Confidence grows with each block on top. This is why exchanges wait for confirmations
- Deliberate waste. The security is the expenditure. There is no way to make proof-of-work cheap without making it weak → energy and externalities
- A 51% assumption. Safety holds while no single party controls a majority of the work (or stake). It’s an economic assumption, not a mathematical guarantee → consensus
Key insight
Blockchains are not a cryptography breakthrough — every primitive they use predates them by a decade. The contribution is an economic one: making the right to write to a shared log costly enough that lying about history is more expensive than the profit from lying. Security here is a budget, not a proof.
Related
- consensus — how proof-of-work and proof-of-stake actually do this
- cryptographic primitives — the parts that already existed
- consensus and Paxos — the classical version, with known membership
- what is cryptography
Source: [reference] — from Nakamoto (2008) and the pre-history. Aug 2026.
Software engineering
Infrastructure & security
Computer science
Hardware & engineering