Two things we build
that cannot be bought.
Data can be bought. Compute can be rented. Models are published every week.
What cannot be bought is accumulated understanding, and the discipline to know what to do when that understanding turns out to be wrong. These are the two foundations of everything we do.
Strategies decay.
Knowledge compounds.
Every dataset we integrate, every label we define, every hypothesis we test and every attempt that fails is recorded with its assumptions and its result. Over time this becomes a structured map of the market: how companies, products, suppliers, people and events connect; which relationships transmit shocks and which only look like they do; which ideas have been tried, under what conditions, and why they stopped working.
This is what makes iteration fast. A new line of research does not start from nothing; it starts from everything the system has already learned, including its failures. Agents query the knowledge base before proposing a hypothesis, and they write back to it after every test. A single strategy may last months. The knowledge that produced it keeps growing, so each round of research starts from a higher point than the last.
Plan for the day it stops working.
Every strategy will eventually fail. The only questions are when, how, and whether we saw it coming. So before a strategy reaches the market, we decide in advance what would prove it wrong, how much we are willing to lose finding out, and what we will do when it happens. Limits on exposure, drawdown and concentration are set before deployment, not after a bad day.
We draw on crises from every market and era we can study, so that events which have never happened in one market have already been rehearsed using another. Live strategies are monitored for drift between what research expected and what the market is doing. Risk is not a brake on research. It is what allows research to reach the market at all, which is why the most important decisions about a strategy are made before it trades.
Models propose. Agents challenge.
The market decides.
Behind every strategy is a large piece of systems engineering that spans quantitative research and AI. Models surface anomalies and turn them into hypotheses. A council of agents challenges each one: questioning its assumptions, searching history and other markets for counterexamples, auditing its data, and testing what it would add to the risk we already hold. Survivors face replication, out-of-sample tests and stress tests inside a fixed risk boundary. Only then does the market give its answer, and that answer flows back into the knowledge base. Most ideas stop somewhere along the way, and every one of them leaves something behind.
Illustrative simulation counts — open the run log
The disciplines behind it
Three disciplines meet inside the system: quantitative engineering, AI engineering, and infrastructure and risk.
Choose any part of the system