Research

What we study, and what we have found.

Five areas. Four are backed by research notes: write-ups of our own work that give the mechanism, the numbers, the sources and the limits. The fifth is collaborative work with other research organisations.

01

Redeemable assets

Staked tokens, vault shares and tokenised claims redeem for an underlying asset after a delay. We study how to price that delay, and how a pool can sell an instant exit without ever selling below redemption value.

  • A floor for the fee, set by the cost of capital over the queue: about 1.3 basis points for seven days of HYPE in mid-2025
  • Queue timing, borrowing rates and the share of holders who redeem, as inputs to the fee
  • Pools that let a provider leave while part of the position is still queued

02

Stale prices, intents and MEV

An automated market maker with an old price pays arbitrageurs to correct it. In intent-based trading, users state an outcome and solvers compete to deliver it. We research where the two meet: how solvers price and route, how value moves through transaction ordering on Ethereum, and how a pool can use solver flow to keep its own price current.

  • HOT: signed quotes that carry the pool’s next price, and a fee that rises with the age of the last update
  • Loss-versus-rebalancing as the measure of what stale quotes cost liquidity providers

Not written up on this site: solver competition and quote design, studied alongside CoW Swap and independent solver teams.

03

Hyperliquid

Hyperliquid runs an order book (HyperCore) and an EVM (HyperEVM) on one chain. CoreWriter lets a contract on one act on the other, asynchronously. We study what can safely be automated across the two, and do data research on Hyperliquid’s markets.

  • Account and permission models for software that acts through CoreWriter
  • Staking, lending and redemption data for HYPE staking tokens

Not written up on this site: fee spend and microstructure on perpetual markets.

04

Exchange architecture

Different assets need different market rules, and most of an exchange has nothing to do with those rules. We design from the asset outward: its redemption process, its settlement delay, the information available when it trades. The parts that do not change are kept in a shared core.

  • A pool core for custody, accounting and the sequence of a swap
  • Separate modules for pricing, fees, access checks, recorded data and reserve storage
  • Two unlike markets on one core: HOT and the liquid staking marketplace

05

Verified and provable code

Two collaborations with research organisations that were building new technology stacks, joined early in each case.

  • With Nethermind: a system that compiles EVM code into Lean, so that properties of a contract can be stated and proved. The aim is formal verification that working developers can use
  • For Starknet, before the lab was incorporated: members of the team built on Cairo v0 and helped mature the language. A Cairo program produces a proof that it ran correctly, which anyone can check without running it again

Research notes

5 notes

Each note was written for this site in November 2026 from research the lab did earlier, and says when that work was done.

Papers

Published elsewhere