It seems to have extraordinary insight into every facet of business and startups that I can imagine. As a team, I'm the better one at inventing a niche business model and writing software in the context of the whole technical architecture, but my relative shortcomings are glaring. In fact it seems like everything I don't know much about GPT-4 knows practically all there is to know! So far it has successfully helped me draft a specific business model in terms of discrete MVP features, narrowing down the whole problem space into the small set of things that can be demoed in about 1 month. I am confident that GPT-4 will be just as capable as we bring our startup into month 2.
If using an LLM as a co-founder is not completely feasible today surely it will be soon. We're clearly much closer to it being possible than impossible. What will this mean for the startup economy? You can imagine a world with more solo founders, less co-founder disputes, a world where founders need to only have expert-level specialty in 1 area because LLMs will fill in the rest of the knowledge gaps.
On the whole it looks like the cost of turning brilliant ideas into tangible products is quickly falling.