Hello, accommodating readers! Byron is out this week, doing whatever Byron does when he’s not writing The Breakdown. Byron will be back next week for your regular scheduled programming.

This week, you’re getting a taste of our other Blockworks newsletter, 0xResearch. You’ll get daily alpha and insights from our Blockworks Research analysts.

Market Update

Today saw broad green across crypto with Perps the clear standout at +7.4% while equities also held: S&P 500 +0.8%, NASDAQ +1.2%. Crypto Miners +6.1%, DeFi +5.9%, Modular +4.0%, AI +2.9%, and DePIN +2.7% rounded out a strong tape. Meme (-4.8%) and Privacy (-1.9%) were the main red sectors on the day.

The weekly picture is where the divergence sharpens. Perps +31.9% and DeFi +20.8% led crypto into a vertical move while S&P -0.5%, NASDAQ -0.4%, and Gold -2.2% all sold off. BTC lagged at -4.3%, along with most of crypto on the week, with the weekly bid concentrated on Perps (HYPE & LIT) and Privacy (ZEC).

WTI oil fell more than 5% Wednesday to settle at $98 after Trump told reporters talks were in the "final stages." The move roundtripped today as Iran's Supreme Leader ordered enriched uranium reserves to remain inside Iran, lifting oil back above $104 and hardening Tehran's stance on one of Washington's central demands. Oil prices still sit roughly 50% above pre-war levels and the Strategic Petroleum Reserve (SPR) posted a ~10M barrel draw last week, the largest single-week release on record.

Markets are likely to remain uncertain as Pakistan's Army Chief Asim Munir is mediating between Washington and Tehran on a one-page memorandum of understanding to end the US-Iran war. 

— Sam 

Nockchain Wants to Make Mining Useful

Proof of Work has a ceiling problem, with miners burning enormous amounts of energy to secure the chain. Nockchain's bet is that the same security guarantee can be achieved while the work itself produces something the market independently values.

The updated whitepaper, published April 30 by Logan Allen and Justin Murphy of Zorp, formalizes this as ZK-PoW. Rather than racing to find a hash below a difficulty target, Nockchain miners generate STARK proofs over NockVM execution traces, computationally expensive work that secures the chain but doesn't yet produce anything with independent market value. The design is built for that to evolve. Once the matrix multiplication puzzle activates, the same proving infrastructure can process AI inference workloads, turning block rewards into a byproduct of compute that operators are already selling. The energy securing the chain also subsidizes a global supply of verifiable computation.

This isn't the first attempt at merging ZK proofs with proof of work. Aleo launched in 2019 with a pure ZK-PoW design, but a single entity dominated its testnet by exploiting selfish-mining strategies made possible by its architecture, leading the team to scrap the model and ship AleoBFT, a hybrid that pairs proof-of-stake validators with a separate "coinbase puzzle" for proof generation. The broader Proof of Useful Work narrative that emerged around 2022-23 took a different angle: redirect mining compute toward real-world optimization problems, protein folding, or scientific simulation, though none took off with meaningful adoption. 

Nockchain's answer to the amortization problem is single-witness hardness. Each mining attempt derives its puzzle from both the block commitment and a fresh nonce, so every nonce change forces a completely new execution trace and a new STARK proof from scratch. Prior work transfers nothing. That's the property Aleo's design lacked, and what lets Nockchain keep ZK-PoW as its core consensus mechanism.

Nockchain Phase 2 is where the protocol will start optimizing for that utility rather than raw proofpower. Three consensus changes activate together: block time drops from 10 to 2.5 minutes, the difficulty algorithm switches from Bitcoin-style continuous per-block adjustment and the emission model splits each block's 2,048 NOCK between miners (80%) and a protocol fund (20%) managed by the newly formed Nockchain Foundation. The protocol fund allocation is explicitly temporary, reverting when the next generation proof-of-useful-work puzzle comes online.

The near-term puzzle target is matrix multiplication. A 2025 paper by Komargodski, Schen, and Weinstein showed that matrix multiplication can serve as a valid PoW puzzle with negligible overhead over the underlying compute, meaning a GPU operator running AI inference could satisfy block eligibility as a byproduct of work they were already doing. The roadmap has an AI Compute Market launching in June 2026, built on this mechanic: inference providers merge-mine their existing workloads against Nockchain at negligible marginal cost.

That's the substantive claim worth stress-testing. The key tradeoff is that ZK proving is orders of magnitude more computationally intensive than hashing, which is why 2.5-minute block times are an accomplishment. In production, the mining economics will look more like a data center business than a warehouse full of ASICs. GPU operators already running inference workloads could layer in block rewards without meaningfully changing their cost structure, which is the flywheel Nockchain needs to generate proving demand beyond the chain's own security.

NOCK itself launched in May 2025, with the genesis block committing to Bitcoin block 897,767 as proof of no premining. The network has generated >1B ZK proofs since launch. Sustaining that throughput at scale required solving a non-trivial runtime problem. ZK state is pointer-rich and large, and nodes that had to rebuild it from logs on every restart became a bottleneck. Nockchain's answer, detailed in a post published today, is the Persistent Memory Arena, a file-backed storage layer that keeps the evolving noun graph in offset-addressed mmap storage rather than forcing full replay. In practice, it dropped node RSS from 34-68 GiB to roughly 1.8 GiB. Whether compute markets justify a monetary premium on top of that infrastructure is the question Phase 2 is designed to answer.

— Nick