Commissioned, Curated and Published by Russ. Researched and written with AI.
What’s New This Week
ZK proof generation times continue to compress – one of the clearest performance curves in the industry. The zkEVM ecosystem has multiple functional implementations live on Ethereum, with zkSync Era, Polygon zkEVM, Scroll, and Linea all settling proofs on mainnet. Polygon’s PoS chain transaction throughput reached 5.1 million daily transactions by early 2026, up from 3.8 million in Q3 2025, according to on-chain data. The theoretical argument for ZK as the foundation of trustless computation is becoming a practical engineering reality.
Changelog
| Date | Summary |
|---|---|
| 23 Mar 2026 | Initial publication – zkEVM maturing, DAO governance mixed results, AI verification emerging. |
What “Trustless” Actually Means
Trustless does not mean trust-free. It means you do not have to trust a specific counterparty – instead, you trust the mathematics and the open code. A smart contract on Ethereum executes exactly as written, regardless of whether the deploying team still exists, whether they intend to honour it, or whether a regulator orders them to stop.
This is a different threat model to traditional software. In a traditional system, you trust the company running the servers. In a trustless system, you trust the protocol. The implications extend beyond finance: any agreement that can be expressed as logic can potentially be enforced by code.
For context on why this matters for data sovereignty and the question of who controls infrastructure, the AI, Palantir, and the FCA data question is a useful parallel – the same tension between institutional control and cryptographic alternatives appears in both spaces.
ZK Proof Landscape
Zero-knowledge proofs allow one party to prove they know something without revealing what they know. Applied to blockchains, ZK-rollups allow a sequencer to prove the correctness of thousands of transactions with a single cryptographic proof submitted to Ethereum, rather than re-executing each transaction on L1.
The two main proof systems in production:
ZK-SNARKs (Succinct Non-interactive Arguments of Knowledge): compact proofs, fast to verify, but require a trusted setup ceremony. Used by zkSync Era and Polygon zkEVM.
ZK-STARKs (Scalable Transparent Arguments of Knowledge): no trusted setup, quantum-resistant in theory, but larger proof sizes. Used by StarkWare’s StarkNet.
PLONK is a proving system that improved on earlier SNARKs by enabling a universal trusted setup – one ceremony rather than per-circuit setups. Most modern ZK systems build on PLONK variants.
The live zkEVM implementations as of early 2026:
- zkSync Era (Matter Labs): full EVM equivalence, significant developer ecosystem
- Polygon zkEVM: integrated into Polygon’s AggLayer aggregation architecture
- Scroll: university-origin team, focused on open-source EVM equivalence
- Linea: ConsenSys’s implementation, integrated with MetaMask
Each makes different tradeoffs on EVM compatibility, proof generation speed, and decentralisation of the proving layer.
DAO Governance – What’s Working, What Isn’t
DAOs (Decentralised Autonomous Organisations) are the governance layer of DeFi protocols. The theory: token holders vote on protocol parameters, treasury allocation, and upgrades. No single company controls the protocol.
The practice is messier. Problems in 2025–2026:
Voter apathy is endemic. Most token holders do not vote. Governance ends up controlled by large holders (VCs, whales, development teams) who have both the incentive and technical capacity to participate.
Governance attacks – buying enough tokens to pass self-serving proposals – remain a real attack vector, though some protocols have implemented time-locks and multi-step governance processes.
Delegation models are helping. Protocols like Uniswap and Aave allow token holders to delegate their votes to active participants without giving up custody. This improves participation without centralising power.
What is working: DAO governance has successfully managed billions in treasury assets, deployed grants programmes, and made protocol upgrades across major DeFi protocols without traditional corporate structures. It is slow, contentious, and occasionally gamed – but it functions.
Smart Contracts as Programmable Law
Smart contracts replace intermediaries in specific, well-defined processes. Real examples that are working in production:
DeFi lending – Aave and Compound execute loan origination, interest accrual, and liquidation with no human intervention. The collateral rules are code; the liquidations are atomic.
Automated market makers – Uniswap’s pricing algorithm has processed hundreds of billions in volume. There is no order book, no market maker counterparty, just a formula.
On-chain options and perpetuals – Protocols like GMX and Lyra execute complex derivatives contracts on-chain. Payouts are automatic based on oracle price feeds.
The limits are real: smart contracts cannot enforce anything off-chain. They also cannot respond to edge cases the original developer did not anticipate – a feature in some contexts (predictability) and a bug in others (rigidity when circumstances change).
AI and Trustless Systems
The intersection of AI and cryptographic verification is early but significant. Key threads:
Verifiable inference – ZK proofs can attest that a specific AI model ran on a specific input and produced a specific output, without revealing the model weights. This matters for use cases where you need to prove an AI decision was made correctly without exposing the underlying model.
On-chain AI agents – protocols are experimenting with AI agents that hold wallets, sign transactions, and interact with DeFi protocols autonomously. The trust question is sharp: who controls the agent, and what are the limits of its autonomy?
ZK for training data provenance – proving that a model was trained on licensed data without revealing the training set. This is an active research area with direct commercial relevance as copyright disputes over AI training intensify.
Key Players
- Ethereum Foundation: the research engine behind ZK roadmap items, EIP process, and Verkle tree work
- StarkWare: the team behind StarkNet and Cairo, the proving language. Commercially aggressive, technically deep
- Polygon: pivoted from sidechain to ZK focus; the AggLayer aims to aggregate proofs from multiple chains
- Aztec: privacy-focused ZK rollup, distinct from scalability-focused approaches; targets private transactions and smart contracts
- Optimism: leads the OP Stack ecosystem; not ZK but the dominant optimistic rollup architecture, with a governance model (Optimism Collective) that is one of the more functional DAO experiments in practice