In today's webinar, the presenter mentioned the importance of using multiple LLMs (ChatGPT, Gemini, Claude) to reduce the risk of hallucinations, errors, and other poor outcomes.
I wanted to mention one tool I've found particularly useful: llm-council. It's an open-source tool that asks multiple AI models the same question, has them review and rank each other's answers, and then uses a "chair" model to combine the best ideas into a single response.
It's essentially a structured way to get a smarter, more reliable answer than you'd get from any one model alone.
I thought I'd share because it has really improved my work and decision-making. I've even used it to design a new diet and exercise program.
Code is available here: GitHub - karpathy/llm-council
If you have any questions, first ask your AI-then feel free to message me or reply here
------------------------------
Jason Bradfield
Past President, San Francisco Chapter
------------------------------