This works great for small sites/apps, but really starts to fall apart if/when it gains complexity where React starts to make sense. I've tried a few times to "just use plain javascript" with the LLM and initial results are often much better, but if the site grows a bit too complex, the LLM starts making a lot of mistakes and it can be hard to reason about as a human and get it on the right track. That hasn't been the case with the React apps IME.