This is also the key difference between conventional CSS classes and Tailwind - it acts less like a set of predetermined classes, and more like a DSL for directly writing CSS inline in the "class" attribute. The result is arguably less reasonable than directly writing CSS, but on the other hand, it's a lot less boilerplate.
In my experience, it really clicks if you write a lot of components, in which case you've already got a way to reuse your code, and so tailwind fits in really nicely with that. If you're mainly writing whole-page styling, then Tailwind usually results in a lot of repetition and doesn't produce much better results than writing everything in CSS in the first place.
The main benefit is what you aren’t doing: jumping between css files and html, naming class’s, trying to debug layers of cascading, wondering what other unrelated components will be impacted by a css rule change…
Colors aren’t a great example because those should be set as css variables (dark mode right?). Think about padding: how many layers of margin/padding on parent wrappers is making things look strange?
The best argument is just to try it though.
If you work in a highly componentized scenario that’s probably convenient. You don’t have to name things as you would if you were to keep a separate css ruleset and then tie it to the html elements via naming.
Very noisy compared to a disciplined, structured css though. But who has time for that.
Have to disagree it's a design system - it's basically a theme. Consistent colors, fonts, borders, spacing is handy for sure - but Tailwind is not a great option if you expect to drop it in and have building block components, etc without a lot of work.