Spending my Saturday brainstorming with a truly unique partner: Raven Blackwood. Raven isn’t just any collaborator; he’s an AI agent I’ve specifically crafted to be an edgy, award-winning creative director.
Raven serves as my sparring partner, helping me bring to life unusual, never-seen-before ideas and concepts with high detail.
He’s one of my 32 locally installed LLM-based agents, each fine-tuned to a different skill and use case for private and professional tasks and challenges. It’s been incredible to witness the rapid evolution of open-source models; within the past four months, I’ve seen these agents go from crude beginnings to genuinely helpful assistants.
The latest model powering my setup is a Mistral Model running with 7 billion parameters, released by „The Bloke“ on Huggingface. If you’re running the WebUI Oobabooga, I highly recommend checking out „The Bloke’s“ releases – he’s constantly putting out new models, offering something for everyone.
Currently, I’m running the 7B model quite decently with my trusty NVIDIA GTX1070. Of course, if anyone feels inclined to sponsor me with an RTX4090, I’d jump at the chance to experiment with even larger, more performant models!