Our Zcash report is now unlocked to the public. Zcash sits at the intersection of expanding financial surveillance and growing demand for private money. The thesis is more relevant now than ever. Read the full report here. delphi.link/Zcash

Can the same computation secure a blockchain and power AI models? Pearl is a Proof-of-Useful-Work blockchain designed to let miners perform AI inference and compete for blocks using the same computation. The goal is to create two complementary revenue streams: block rewards help subsidize inference, while customer payments help cover the underlying cost of mining. Pearl uses matrix multiplication as its mining primitive, which is also a core operation in AI inference. When a miner processes a compatible inference request, the same calculations can generate mining attempts while producing an output that serves a customer. But miners can also perform arbitrary matrix multiplications solely to compete for block rewards. The protocol verifies that the prescribed computation was performed correctly, not whether the matrices originated from a real inference request. The economic incentive is straightforward. Miners performing real inference can earn customer fees and mining rewards from the same underlying work, potentially improving their economics relative to miners performing synthetic calculations alone. The remaining challenge is converting that incentive into realized utility: how much of the network's total computation is actually serving external demand rather than being performed solely to earn block rewards?

Hyperliquid recently crossed $7B in TVL for the first time. The exchange has been seeing substantial activity with about $220B in perp volume and nearly $60M in protocol revenue over the past month. TradeXYZ’s HIP-3 markets alone accounted for nearly a quarter of Hyperliquid’s perp volume over that period.

Year One.

Elysium could become Hyperliquid’s next growth engine. Hyperliquid dominates perp trading, but high gas costs and limited throughput have held back token launches and spot activity on HyperEVM. Kinetiq’s Elysium is designed for faster execution and direct connectivity to HyperCore. A token can launch and build liquidity on Elysium before pursuing a HyperCore spot market and HIP-3 perp. Lower execution costs could make it easier for prop AMMs to support early trading. Once a liquid perp exists, they can hedge their spot exposure on Hyperliquid. If projects launch there and their tokens keep trading, more of that activity could reach Hypercore. That’s the growth case Elysium has to prove.

Derive holds the top three ETH and top two SOL spots on the Delphi options dashboard. The three ETH contracts did ~$200M in 24h notional, roughly 11% of tracked ETH options volume. All are Sept. 24 calls at $2,850, $3,000 and $3,100. On SOL, Derive’s December $150 and $300 calls did another ~$3.45M, roughly 12% of tracked SOL options volume.

Which launchpad is built to win the next memecoin cycle? PONS has a clear revenue model: it shares trading fees with creators and uses part of its share to buy and burn PONS. That model generated $21M in revenue so far in September. But the next cycle will depend on where breakout tokens launch and keep trading. Several of Robinhood Chain’s strongest runners have come from LONG. LONG has no platform token of its own. Each memecoin it launches is paired with a Robinhood Stock Token. In community mode most buy fees go into a permanent vault holding the paired asset. A breakout can therefore build a lasting stock-linked balance in its community vault. LONG has already launched breakout tokens. Now it needs to keep buyers in their paired pools after the initial surge. Sustained demand would grow the community vaults and strengthen its case against PONS.