There are two main optimizations: 1. We have optimized Module.concat/2 (it is roughly 100% faster) 2. We have optimized rendering of Inspect.Algebra by skipping the intermediate s_doc representation. This reduces memory allocation and skips one extra pass on rendering.