HOW WE WORK

Four disciplines,
one small team.

A venture does not need four agencies. It needs these four things to happen in the same room, in the same week, by people who are still there next month.

The four disciplines

The handovers
that never happen.

Four disciplines in one team is not an org chart. It is the absence of four handovers — and a handover is where a venture usually loses a month, along with most of what it had learnt.

RESEARCH → BUILD

A thesis is not a brief

Where it usually goes

A research phase produces a document, the document is handed to whoever builds, and the kill criteria do not survive the journey. Months later the thing exists and nobody can say what it was meant to prove — so it gets judged on whether it works rather than on whether it was worth building.

What happens instead

The people who wrote the kill criteria are the people who build against them. The number in the thesis is the number on the dashboard, with the same definition, not a cousin of it.

BUILD → THE MARKET

The cost model is not an afterthought

Where it usually goes

Engineering optimises what engineering can see: latency, uptime, throughput. The fee, the spread and the failed settlement land on a finance spreadsheet nobody in the build reads. The system comes out fast and the unit economics come out negative, and the two facts are discovered a quarter apart.

What happens instead

What a transaction costs to run is a figure the engineers hold, and it is reviewed in the same meeting as the latency. When it moves the wrong way that is a build problem, not a reporting problem.

PRODUCT → THE RULE

Counsel is not a final inspection

Where it usually goes

A product is designed, shown to a lawyer, and redesigned. The expensive version of this is finding the constraint after the integration is built and the counterparty has signed.

What happens instead

The constraint is in the thesis from the first week and the design starts there. It is a great deal cheaper to design within a rule than to discover one.

LAUNCH → RUNNING IT

Somebody has to be awake on a Sunday

Where it usually goes

Reconciliation breaks, a settlement fails, a key needs rotating, a counterparty quietly changes a field. The work that keeps a money system upright is invisible in a plan and unavoidable within a month of launch.

What happens instead

Operational load is estimated before launch and carried as a cost inside the number, rather than absorbed by whoever happens to be nearest. A venture that only works while a founder never sleeps has not been tested yet.

The rule we hold to
when the numbers look good.

A result that has not been measured net of what it costs to run is not a result. Spreads, fees, failed settlements, the person who has to check something by hand — they come out of the same number the upside goes into. Most promising ideas in these markets die exactly there, and it is cheaper to find that out in week two than in year two.

It is the least popular discipline in the building, because it is the one that takes a good result away. It is also the cheapest thing we do. The cost is the finding works through a case where the entire result was transaction cost, and the signal underneath it was worth nothing at all.