Scaling & the Blockchain Trilemma
A base-layer chain is slow because every node redoes every transaction — so you scale by moving the work off it, not by adding servers.
You used a blockchain and it was slow, and the fee was absurd — meanwhile the laptop in front of you does millions of operations a second. How is a global network of thousands of computers SLOWER than your phone? And why can’t they just “add more servers” the way a normal app does? Hold the question — the answer exposes a deep tradeoff sitting at the very center of crypto.
Why base layers are slow on purpose
Every full node re-executes and stores every transaction to verify the rules itself (lesson 10). So the network’s capacity isn’t the sum of all those computers — it’s capped at what a single ordinary node can handle, because each one has to keep up alone. That’s why base layers are slow and why block space is scarce and fees rise under load (lesson 9).
This isn’t bad engineering; it’s the price of letting everyone verify everything without trusting anyone.
The trilemma: pick your tradeoff
The blockchain trilemma says it’s very hard to maximize all three of decentralization, security, and scalability at once. Crank scalability the obvious way — make blocks huge or lightning-fast — and running a full node gets expensive, so fewer people can run one. Fewer independent nodes means more centralization, which weakens the very trust guarantees that made the chain worth using (lesson 10).
So naive scaling quietly spends decentralization to buy speed. The hard part is scaling without paying that price.
The big-blocks tradeoff: • Want 10× the transactions? Make blocks 10× bigger. • Now every full node needs far more bandwidth, storage, and compute to keep up. • Hobbyists and ordinary users can no longer run a node; only big, well-resourced operators can. • Verification concentrates in a few hands — you got throughput, but the network is now more centralized and easier to pressure. Scalability was bought with decentralization.
Scaling by moving work off the base layer
The winning approach isn’t to make the base layer do more — it’s to do most work elsewhere and use the base layer only as a secure settlement anchor. Layer-2s, especially rollups, execute thousands of transactions off the base layer, then post a compressed summary plus a proof back to it. Every base-layer node verifies one small proof instead of thousands of transactions — cheap — while the result still inherits the base layer’s security.
Honest caveat: layer-2s add their own complexity and trust assumptions (who posts the proofs, whether the data is available to all). So “fast and cheap” still earns the question from earlier lessons: what did it trade away, and can you verify it?
The base-layer chain is a country’s supreme court: supremely trustworthy and final, but slow and costly — you’d never route every parking dispute through it. So most matters are settled in local courts (layer-2s), and only the final outcome, or a genuinely contested case, gets recorded at the top. A rollup “rolls up” thousands of everyday transactions into a single entry the supreme court records — so the court’s authority backs millions of cases without it personally hearing each one.
How a rollup scales without bloating every node: 1. On the base layer, 1,000 token transfers would be 1,000 separate transactions — slow and pricey, since every node processes all 1,000. 2. A rollup executes those 1,000 transfers off the base layer. 3. It posts ONE compressed record back to the base layer: the net result plus a proof that all 1,000 were valid. 4. Every base-layer node verifies that one small proof — cheap — instead of 1,000 transactions, yet the outcome is secured by the base layer. 5. Users got 1,000 transfers of throughput while the base layer did roughly one transaction’s worth of work. Execution moved off-chain; settlement stayed on-chain.
This is the reading. The interactive version — active-recall quiz, a hands-on experiment you run in your own AI, and an earned mastery check — is free in the app.
Start this lesson free →