Yeah, but imagine the first engineers at ChatGPT back in 2021, some thrown together group of people who knew frontend the best from OpenAI, a research lab, to create a quick prototype UI for chatting with a text generating ML model. Of course they'll take some shortcuts, that's to be expected. All of this is just guesses though, but seen similar things play out many times. Of course, at one point someone needs to step in and pause for a moment to plan ahead for a bit, but depending on the leadership, it isn't always so easy to do.

No doubt there’s a lesson in there, but I don’t think it’s a technical one.