Cartesi (CTSI)

In a recent thread, @joaopdgarcia a Cartesi developer explained the difference between fraud proofs and ZK proofs in rollups.

He described fraud proofs as a way to “reveal lies” by re-executing disputed steps, and ZK proofs as mathematical statements asserting correctness without re-execution.

That distinction is useful and I quite agree with him .

ZK proofs are impressive as they compress a computation into a tiny proof that anyone can verify quickly.

But this power comes with trade‑offs.

Surveys like the 2025 SoK on zk‑SNARKs point out that scaling these proofs for real‑world, Turing‑complete computations is challenging.

Circuits can become extremely complex, proving requires significant memory and hardware, and only a few actors can generate proofs efficiently.

Studies comparing SNARKs, STARKs, and Bulletproofs confirm that even seemingly simple computations can be orders of magnitude more costly to prove than to execute.

So while ZK proofs promise instant validity, the practical costs, both computational and economic, are non‑trivial.

Fraud proofs, on the other hand, remain surprisingly powerful in practice.

The idea is simple: assume computations are correct unless challenged.

If a dispute arises, only the specific step in question is re‑executed and verified.

Research on optimistic rollups, including recent 2024–2025 studies, shows that fraud proofs not only ensure correctness, but also preserve decentralization, sequencer honesty, and economic security.

Because anyone can verify a dispute cheaply, there’s no natural bottleneck or centralization risk.

They also allow for full Turing‑complete off‑chain computation, something current ZK circuits cannot reliably handle due to expressiveness limits.

The other dimension is hybrid designs, which combine the strengths of both approaches.

By using ZK proofs where computation is tractable and fraud proofs as a fallback for complex or edge‑case logic, hybrid rollups can reduce prover centralization, maintain high throughput, and strengthen security guarantees.

Recent research from 2025 shows that hybrid architectures can balance finality speed, computational expressiveness, and decentralization in ways that pure ZK or pure fraud‑proof systems cannot.

They also provide better transparency for MEV management, since fraud proofs allow visibility during the challenge window, while ZK proofs compress it.

Looking at all the evidence, my take agrees with Joao.

The debate isn’t about choosing one over the other;

it’s about understanding the trade‑offs and designing systems that leverage the strengths of both.

For anyone building Layer‑2 rollups today, this means using fraud proofs for complex logic, applying ZK proofs where proving costs are manageable, and considering hybrid architectures to maximize security, throughput, and decentralization.

Make sure to read the entire thread here 👇