This is not the most cogently written argument ever, but I think the points here are all essentially correct.

The key risk highlighted here which I have not seen as much talk about is the way that these technologies might give shift power from white-collar labor to capital in a drastic way. The ability to build the AI products that people will actually use on a day to day basis seems like something established companies with lots of customer data will be hard to compete with. For example, its pretty trivial for Salesforce to plug a LLM into their products and get pretty good results off the bat.