The author has missed a critical insight from their evaluation — the fact that they could develop a tool in vanilla JS more simply than React is only because they first developed it in React. Developing with a framework gives you the advantage over bespoke tooling — you can experiment and iterate faster. It is only having had that experience, and exploring the problem-domain so clearly by building it once, does the second build happen to quickly.

So yes, maybe React was over-engineered for their problem. But I think they only realised they over-engineered their problem after having over-engineered their problem...

A better point would be "I used React to test-build an app then to build it again natively for speed."

--

As an aside, I do agree that React (and I'm extrapolating this to many of the web frameworks) have an incredibly steep learning-curve and onboarding cost. I'm an avid Vue fan and I've found the architecture around Vue 3's "single-page components" to have created so much technical debt and complexity that I spend more time debugging the build pipeline then actually developing. I still like the flexibility and acceleration I get with Vue, but Vite often leaves me feeling like I'm just fighting the framework... I guess I just miss Gulp.