[Intermediate] — the test that kills most proposals, the cases that survive it, and why that’s still worth learning.
The test
A blockchain is a slow, expensive, public database with no administrator. You accept all of the first three to get the fourth. So:
Is there any party who could be trusted to run this, and if so, use a database.
That’s the whole test, and it eliminates the large majority of blockchain proposals — including nearly every enterprise pilot from 2016–2020, most supply chain projects, and essentially every “blockchain for X” where X has a natural operator.
The extended version, four questions:
- Do the participants distrust each other? If not, one of them can run the database
- Is there an acceptable administrator? If yes, use them. They’ll be faster, cheaper and fixable
- Does the data originate on-chain? If a human types it in, the chain guarantees the record wasn’t altered, not that it was ever true
- Does anything break if a company shuts down? If not, you didn’t need censorship resistance
What survives — and it’s a real list
1. Censorship-resistant value transfer. Sending money that no intermediary can block. This is the original use case and it remains the strongest.
Concretely useful where the alternative is bad: remittances to countries with expensive or restricted corridors; holding savings under high inflation or capital controls — Argentina, Turkey, Nigeria, Venezuela have real, non-speculative stablecoin usage, driven by currency collapse rather than ideology; and payments to people excluded from the banking system for political reasons.
This is not a hypothetical benefit, and it’s the case Western critiques most often miss because dollar-denominated bank accounts make the problem invisible.
2. Trust-minimised coordination between adversaries. Escrow, settlement and clearing between parties with no basis for trust and no shared jurisdiction. Where a neutral operator either doesn’t exist or would be too powerful.
3. Credibly permanent commitments. Rules that cannot be changed later, provably — an immutable contract, a fixed supply, a distribution nobody can alter. The value is in the impossibility of the change, and no ordinary system can offer that, because “we promise not to” is always revocable.
4. Permissionless composability. Anyone can build on a deployed contract without asking. No API keys, no terms of service, no revocation. This is genuinely novel and underrated — it’s why DeFi assembled as fast as it did, and there’s no equivalent in ordinary software → composability.
5. Verifiable computation. The zk proof machinery is a genuine primitive that didn’t exist before — proving a computation was performed correctly, cheaply, without redoing it. Its applications extend well beyond blockchains → zk proofs.
6. Public, permanent registries where the entry is the asset. ENS names, on-chain art, DAO membership. Not pointers to things elsewhere — the record itself → ENS.
What doesn’t survive, and why
Supply chain provenance. The chain preserves what was written, not whether it was true. Someone scans a QR onto a pallet, and a lie is now permanent and tamper-proof. Garbage in, immutable garbage out → the oracle problem.
Enterprise consortium chains. Known participants, a shared operator, no censorship concern. A replicated database with signed writes is the same thing at a thousandth of the cost, and this is broadly what those projects concluded.
Voting in public elections. A public ledger makes vote-buying verifiable, which is worse than the problem it solves. Secret ballots exist for good reasons, and the failure modes of election systems are overwhelmingly about identity, coercion and endpoint security — none of which a chain addresses.
Most identity. Needs a trusted issuer, which is the thing being avoided → identity.
Storing files. Wildly more expensive than any alternative, for content addressing you can get from a hash → storage.
Anything needing speed, privacy, or the ability to fix a mistake.
The honest summary
Blockchains are a narrow, expensive tool that does one thing nothing else does: maintain agreed state among parties who don’t trust each other, without an administrator.
That is genuinely valuable in a small number of situations, and the field has spent enormous energy pretending it’s valuable in a large number. Most of the value created has been financial — trading, speculation and the infrastructure serving them — plus a real, under-reported base of people using stablecoins because their own currency failed them.
A blockchain in an application where a database would do isn’t neutral — it’s strictly worse: slower, costlier, publicly leaking your data, and unfixable when wrong.
Why learn it anyway
Because it’s a genuinely interesting piece of distributed systems engineering, and because the surviving use cases are real. Also:
- The engineering transfers. Adversarial thinking, invariant-based design, and building where you cannot patch make you better at everything else
- The economics are novel. Mechanism design where the code enforces incentives is a real discipline, applicable well beyond chains
- The cryptography is real. zk proofs and threshold signatures matter regardless of what happens to crypto markets
- It employs people well, and knowing it properly — including its limits — makes you far more useful than the average practitioner
Learn it as a distributed systems specialisation with a strong adversarial security component. That framing is accurate, it survives market cycles, and it’s what the good engineers in the field actually do.
Key insight
Ask “who would I have to trust, and can I?” — if there’s an acceptable answer, you don’t need a blockchain. That question resolves nearly every proposal in seconds, and being the person in the room who asks it is more valuable than being the person who can write Solidity.
Related
- what web3 actually is — the same argument, at the start
- the fraud surface
- gaming and the rest — the four-question test applied
- distributed systems — the field this belongs to
Software engineering
Infrastructure & security
Computer science
Hardware & engineering