Simon Willison quotes Molly White's argument that large language models may only offer small productivity gains because they're only somewhat useful for mundane tasks. via @simonw