Angularjs is reasonaly good library but as many other big frameworks it tries to absorb every role in itself. This leads to code bloat of the framework and often limits developers with API available.

This problem repeats in angularjs on different levels. The top manifistation is module system.

Authors defined notion of module and registry for controllers, providers, directives, and so on. This is an example from Angularjs doc:

var myModule = angular.module('myModule', []);

// add some directives and services
myModule.service('myService', ...);
myModule.directive('myDirective', ...);

This API directs developers to build angularjs centric applications, because it's where module and registry is defined.

At the same time there exists de-facto standard API called Asynchronous Module Definition (AMD) to specify a mechanism for defining modules and their dependencies. This API has several implementations, where requirejs is, probably, the best known one. This is an AMD example:

define(["./cart", "./inventory"], function(cart, inventory) {
  return {
    color: "blue",
    size: "large",
    addToCart: function() {
      inventory.decrement(this);
      cart.add(this);
    }
  }
});

Every angularjs artifact can be mapped to AMD module. This could work equally well both at run time, and during unit tests. Without custom module implementation angularjs would be smaller, and more modular, so developer could pick up only required components. At the same time with AMD angularjs could treat many existing javascript classes as controllers and services without specific adaptation. Again, being more modular angularjs could provide multiple implementations of the same feature like jqLite vs jQuery, or $q vs native promises, for a developer to select.

So, we think modules in angularjs is a flaw, whose correction would improve it in many ways.