Ethereum

Ethereum is a blockchain whose distinguishing feature is a general-purpose virtual machine. Where Bitcoin’s scripting is deliberately limited, the Ethereum Virtual Machine executes arbitrary programs — smart contracts — with their state stored on-chain and their execution replicated by every node.

That replication is the whole design and the source of both properties worth caring about. Because every node runs the same computation and agrees on the result, no single operator can alter it. And because every node runs the same computation, it is extraordinarily expensive relative to running it once on a server.

What the guarantee actually covers

The guarantee is narrow and worth stating precisely: the code deployed will execute as written, and its state cannot be changed except through the rules the code allows.

It does not guarantee the code does what anyone intended. Contract bugs are permanent by construction — the same immutability protecting against tampering prevents fixing a mistake. Nor does it say anything about data arriving from outside the chain, which is the more fundamental limit.

A blockchain can only verify what it can compute. Anything about the external world — a price, a delivery, a measurement — has to be reported by someone, and at that point the trust assumption moves to that reporter. Oracles are the mechanism, and they mean a system marketed as trustless is usually trust-shifted rather than trust-free. This is the part I find most often glossed over.

Why it is a small note here

Ethereum sits at the edge of this garden, and I want to be straightforward about why. The technical design is genuinely interesting — consensus mechanisms, the economics of gas as a computation price, the proof-of-stake transition — and the applications I have found convincing are considerably narrower than the surrounding claims.

For the research data problems I work on, the questions blockchain is offered as an answer to — provenance, integrity, attribution — are better served by provenance metadata, persistent identifiers, and institutional commitments to maintain them. Those solve the actual problem, which is usually that nobody recorded what happened, rather than that someone might tamper with the record afterwards.

See also: smart contracts and the computing map.