Research Engine

What it is

The Research Engine is not a product anyone buys. It is the internal machinery that moves an idea from a hypothesis about market structure to a modelled result, to a reference implementation, to something that holds real capital — and, more often, kills it somewhere along the way.

We describe it publicly because the sequence is the thing that distinguishes us, and because every other product on this site came through it.

Simulation

Most of the work is simulation against real chain state. A mechanism that is elegant under an assumed order flow is worth very little; the question is always how it behaves against the flow that actually exists, including the adversarial part of it. Replaying historical state against a candidate mechanism is how a design earns the right to a testnet.

Reference implementations

Every primitive gets a reference implementation that is deliberately not optimised. It exists to be read, argued with and forked. Where we can, we open-source it, on the reasoning that a primitive nobody can inspect is not really infrastructure.

Publishing

Findings are published whether or not they support a product we are building. Several of the lab’s papers describe mechanisms we tested and rejected, which we think is the more useful half of a research programme.