how to evolve a system

whatever a system tests is what it becomes. every system evolves through three mechanisms: continual randomised novelty, tests that favour some of it, and spread of the winners. set the tests deliberately and the system marches toward your goal; leave them to chaos and it still marches, just not for you.

novelty

all innovation arrives inside novelty. a system improves by raising the volume of mutations or the ratio of good ones: toyota harvests ideas from every factory floor because discarding a bad one costs nothing, while genentech aims each new treatment at a known biomarker. volume or quality, the mutation supply sets the ceiling.

tests

the tests imposed on a system decide the ideals it evolves towards. one survival test made peppered moths light before the industrial revolution and dark after it, and the nba quietly selects giants from a population where almost nobody stands that tall. be exact about the test and the system does the rest.

spread

a winning trait dies with its entity unless it spreads. the fosbury flop remade the high jump because every jumper who saw it could copy it; fashion only cycles, because status is zero-sum and cannot raise the whole system. spread converts one entity’s luck into the system’s standard.

speed

evolution accelerates at its bottleneck. once the loop exists, speed comes from one mechanism at a time: produce more or better novelty, test it faster and at volume, or cut the friction slowing the winners’ spread. growth, above all else.