Joe Weisenthal
@TheStalwart
One half of Bloomberg's Odd Lots Podcast. One quarter of Light Sweet Crude.
加入 March 2008
7.2K 正在關注 450.4K 粉絲
Making some progress on getting the model to dispatch sub-agents to compete on generating new ideas and then having GPT6 judge the idea. No idea if it's working, but it's kind of interesting to try observe.
顯示更多
Where’s a guide to getting a model to just keep trying various techniques to achieve some goal. I have two experiments I’m working on that are slightly more complex than @Havelock_AI / Fedlock. And I want the AI to just autonomously cycle through new approaches until one works.
顯示更多
