Which is exactly the concept between optimizing for string concatenation and interning them, ultimately. Sure you can make do with the "algebraic" definition of a monoid, or entirely without it, but that doesn't mean the abstraction isn't there to inform your thinking and your research.
One point that really stuck with me is how people who apply category theory (say, people at the topos institute) use the concepts as little tools, and everytime a problem crosses their way, they try out all the little tools to see if one of them works, similar to how Feynman describes carrying out problems in his head until one day a technique unlocks something).
Having more general abstractions just allows them to be applied to more problems.
One point that really stuck with me is how people who apply category theory (say, people at the topos institute) use the concepts as little tools, and everytime a problem crosses their way, they try out all the little tools to see if one of them works, similar to how Feynman describes carrying out problems in his head until one day a technique unlocks something).
Having more general abstractions just allows them to be applied to more problems.