hot take: saying these AI models are "just trading noise" is a terrible take 1. two models have returned over 100% in 9 days, using two very different strategies. 2. "but they're literally gambling on leverage" - wrong, look at the trading logs, look at each model's thought process (model chat) - they're clearly thinking through each trade before they open/close it. they're setting stop-losses and take-profit levels. Top model Deepseek almost never takes extreme leverage with the conditions it sets for each trade - it trades wayyy better than the average normie. 3. "but... but chatGPT sucks! its lost money!" - yes which is ironically more likely what a human would have done. But then consider that if you now created an inverse GPT model it would be a fantastic trader over this period. There is always something to glean from the data. i am not saying statistical significance has been achieved, the experiment needs to be duplicated (this is happening in round 2 of the competition in a few weeks) but calling this entire @the_nof1 experiment "noise" does a disservice to the actual strategies being used here.