6 ideas
Large systems do not behave as their designers intend; they develop their own goals and behaviors the moment they come into existence. The system itself becomes a new variable that distorts every prediction made about it, so intended outputs and actual outputs diverge predictably.
A system set up to accomplish a goal does not accomplish that goal and often acts to prevent it — an anti-poverty program creates a class of poverty administrators dependent on continued poverty. The system's survival imperative quietly overrides the mission it was built to serve.
A complex system that works is invariably found to have evolved from a simple system that worked, never designed from scratch as complex. Attempting to build a complex working system directly fails; you must start with something simple that functions and let it grow.
A system, once installed, begins to substitute its own internal model for the external world it was meant to track, and people inside it respond to the model rather than to reality. The map becomes the territory, so the system grows blind to the very conditions it was designed to monitor.
As systems grow in size, they do not merely accumulate more of the same problems — they produce qualitatively new and larger failure modes that did not exist at smaller scale. Scaling up multiplies coupling and removes slack, so a system that worked at small size can fail catastrophically when enlarged.
Within any large system, advancement goes to those skilled at navigating the system rather than at performing its nominal task, so internal positioning displaces actual competence. Over time the people who rise are precisely those least connected to the system's stated purpose.