I think wrapper companies could do very well if AI does end up becoming having an impact near on the order of the internet or smartphones, which it may never.

I think wrapper companies could do very well if AI does end up becoming having an impact near on the order of the internet or smartphones, which it may never. A product that incorporates a foundation model could still require solving a hard engineering problem separately from the training and provision of said model. Someone who could, say, provide a reliable filter for copyrighted content from an LLM is providing real value and has a tangible moat.