[lewq]

Vision language models have been trained on how to operate human UIs though, so at least for a while, computer use will be an interesting area to explore. I think debugging web apps and building UIs is a particularly fruitful area for this