Features are copied. Loops are not. A feature is a point in a product. A loop is a system in which each use of the product makes the product more valuable, more defensible, or more efficient than it was before. Features compete on execution. Loops compete on gravity. The companies that become generational are the ones that figured out, early, which loops were available to them and invested in those loops at the expense of feature parity. The companies that plateau are the ones that spent their years shipping features their competitors shipped three months later.
The non-obvious insight is that loops are usually invisible in the product itself. The loop is not the feature the user sees. The loop is the structural consequence of the feature being used. Every search on Google improves the search index. Every ride on Uber improves the liquidity model. Every review on Amazon improves the conversion rate for the next buyer. None of these are features a user would point to. They are the consequences of features, and the consequences compound. A founder who understands their loops can tolerate feature inferiority for years, because they know that each week of use is widening the moat regardless of what the competitor ships.
Founders get this wrong when they benchmark themselves feature by feature against competitors. This is a losing game because features are the most copyable surface of the product. Founders get this right when they can draw, on a whiteboard, the specific mechanism by which usage of their product creates an advantage that a competitor cannot replicate by shipping the same features. If they cannot draw it, they do not have it.
Most companies are using AI as a feature. They bolt a model onto an existing workflow, call it AI-powered, and ship. This produces parity, not advantage, because every competitor has access to the same models. The companies that will actually win are using AI to build proprietary loops: capturing a type of data that only their product generates, training or tuning on that data in ways competitors cannot replicate, and improving a workflow that becomes more valuable with every customer. The feature is the same. The loop is not. In five years the gap between the two approaches will be unbridgeable, and most of the AI-feature companies will be footnotes in the AI-loop companies' case studies.
The idea worth holding onto is that durable advantage almost never comes from being smarter or working harder than the competition. It comes from building a structure in which each new user, each new piece of data, each new transaction makes the next one more valuable, so that the company is getting stronger while its founders sleep. Network effects, data loops, and supply-side economies of scale are not marketing concepts. They are the physics of why some companies become unassailable and others, equally well-run, remain vulnerable forever. This belongs under the sixth rule because founders are drawn to surface cleverness: the feature that wins this quarter, the campaign that gets the press. Those things do not compound. Loops compound. The real work of the first years is to identify what loop the business could plausibly run on and to build toward it, even when the loop is invisible to everyone else and the visible features are getting all the attention.
Signing this rule means you commit to building systems that get harder to compete with over time, not products that get easier to copy.