I think it‘s only fair to be critical since there is a product attached to this experiment

You don't really need an LLM to beat SPY by choosing the largest four holdings of SPY. Plenty of passive funds already do something like that. It's a fun experiment, but don't misinterpret it to mean the LLM had a unique insight.

To frame it another way: When the best returns were from investing in the company that makes the GPUs that run the LLM that you're asking for stock picks from, did you really need the LLM?