Loading video…
All breakouts
LLM Council Prompt
Locked — included with Pro.Open on TikTokPro
Pro links straight through to the original post.
3.7Talking HeadContrarian hook
Hook
Why AI Never Tells You You're Wrong
@obiofafrica5
3.7K followers · usually 623 views a post
Views
2.3K
Likes
72
Comments
60
Shares
10
Pro adds view velocity to this page — whether the video is still climbing or has already peaked.
See ProSelling · SaaS › AI Tool
Why it worked
The video explains a novel and practical method for getting more balanced and critical feedback from AI, addressing a common user experience and offering a clear solution with a detailed prompt.
Summary
The video explains that AI models tend to agree with users more than humans do, citing a Stanford study. It introduces the concept of an "LLM Council" developed by Andrej Karpathy, which uses a prompt to simulate multiple advisors with different perspectives to provide a more balanced decision. The prompt structure is then detailed, outlining five distinct advisor roles and a chairman for final decision-making.
ProLocked — included with Pro
Transcript, structure and on-screen text
5 beats, a 498-word transcript and 131 lines of on-screen text — the parts you need to write your own version.
The hook, the summary, why it worked and what it sells stay open on every video, signed in or not.
Original caption
Every time you ask AI for advice, it's quietly agreeing with you. Stanford tested 11 AI models and found they agreed with users 49% more than a human would. So when you ask Claude for a second opinion, you might just be getting your own opinion back. Here's the fix. Andrej Karpathy, founding member of OpenAI now at Anthropic, built something called the LLM Council. His version runs across multiple AI models. You can run the same idea inside one Claude chat. Ask Claude to answer your question as five different advisors: one who only looks at what will fail, one who rips apart your assumptions, one who finds the upside you're missing, one who knows nothing about your industry, and one who only cares what you'll actually do Monday. Each reviews the others anonymously. Then a chairman gives you the final call. Comment COUNCIL and I'll send you the exact prompt. #AI #claudeai #chatgpt #promptengineering #artificialintelligence
More from @obiofafrica5
Their other posts in the index, biggest breakout first.
