We talked about our decision-making system.

We feel that it’s the most important thing for building a company. If our decision-making system is good, everything else will fall in place. However, it turns out to be very hard to design - and that’s the reason most companies continue operating without one (wasting time on pointless meetings, and wasting resources on pointless products).

Our system is different from others in many crucial ways.

The most important one is that decisions are made by writing code. Every decision is a function that takes an array of elements and returns a single element from that array.

The second one is that decisions are made objectively: all properties of the elements must be defined by the measurements of the actual world (not subjective feelings).

The third one is the use of a stable random estimator in case where it’s impossible to define any objective properties. It’s important for the estimator to be stable (not change its values across multiple runs). This is achieved by fixing a constant seed for the random number generator.

There’s much more to the system than could be explained in a single AMA. We plan to publish more information about it in future.