Simple Rules. Big Effects.
A flock of starlings can produce one of the most complex visual patterns in nature. The rules governing it fit on an index card. Three rules. The complexity emerges from the interaction.
A flock of starlings can produce one of the most complex visual patterns in nature. Thousands of birds, moving in perfect coordination, creating shapes that look choreographed by a mind far larger than any single bird possesses.
The rules governing this behavior fit on an index card. Stay close to your neighbors. Match their speed. Don't collide. Three rules. The complexity emerges from the interaction, not the instructions.
I keep coming back to this principle because it shows up everywhere I look.
Design systems
The best design systems aren't the thickest ones
I've worked with brand books that ran to two hundred pages and brand books that ran to thirty. The thirty-page systems produced better, more consistent output. Because the constraints were sharp enough that any competent person could apply them without interpretation. The two-hundred-page systems produced meetings. Meetings about what the system actually meant, which section applied, whether this exception was covered somewhere in the appendix.
More rules didn't produce more consistency. Fewer, more precise rules did.
Code
Code works the same way
The clever solution: the one that compresses fifty lines into twelve using tricks only the author understands. It's worse than the readable version that takes up more space but can be maintained by anyone on the team. Readability is a simple rule. It beats cleverness in every codebase that lasts longer than six months.
Business strategy follows the same pattern.
The companies I've watched succeed at scale usually had a strategic constraint you could state in a sentence. Not a mission statement. An actual operating boundary that shaped every decision. "We only serve teams of 50-500" is a simple rule. It eliminates a thousand decisions about features, pricing, support, and go-to-market. Every decision that doesn't need to be made is friction removed from the system.
AI
AI is where this principle becomes most visible, and most violated
The dominant approach to AI right now is more. More parameters, more data, more compute, more tokens in the context window. And it works, in the sense that the output is impressive and the demos are convincing. But impressive output from an opaque system is not the same as reliable output from a constrained one.
When I build AI systems, I spend more time on boundaries than on capabilities. What the system won't do matters more than what it can do. A constraint that says "always read from this schema before generating" is a simple rule. It eliminates an entire class of hallucination. A constraint that says "the user confirms before any learning is applied" is a simple rule. It eliminates the black box problem.
The instinct is always to add.
Add a feature, add a parameter, add a fallback, add an exception. But the systems that work, really work, in production, at scale, over time, are usually the ones where someone had the discipline to subtract.
Simplicity isn't the absence of complexity.
It's the result of understanding a problem well enough to find the few rules that generate the behavior you want. The starlings don't need a choreographer. They need three rules and the space to fly.
The hard part was never building complex systems.
The hard part is finding the simple rules that make complexity unnecessary.