Beyond the Echo Chamber: Dan Sherrard-Smith on Using AI for Genuine Feedback

In a recent LinkedIn post, Dan Sherrard-Smith discusses a common pitfall founders encounter when using AI tools like Claude: the tendency for the AI to act as an echo chamber, confirming pre-existing beliefs rather than challenging them. Sherrard-Smith argues that this passive affirmation can hinder true innovation and lead founders astray.

The core of his observation is that many users interact with AI in a way that elicits agreement rather than critical evaluation. He highlights this by stating:

“Most founders use Claude to confirm what they already think. They describe an idea. Claude says it’s promising. They feel good and they move on. You are literally paying for a yes man.”

Sherrard-Smith contends that this approach negates the potential of AI as a tool for robust problem-solving and strategic development. He emphasizes that the true value lies not in validation, but in constructive criticism and the identification of potential flaws.

The ‘LLM Council’ Skill: A Framework for Critical AI Interaction

To counteract this tendency, Dan Sherrard-Smith introduces and champions a specific AI skill developed by Ole Lehmann, termed ‘The LLM Council.’ This skill is designed to fundamentally alter how users engage with AI by simulating a panel of advisors with distinct critical perspectives.

According to Sherrard-Smith, the LLM Council moves beyond a single AI response to a more dynamic and challenging interaction. He explains the concept:

“Instead of one Claude agreeing with you, five advisors attack your question from different angles and then peer-review each other before giving a verdict.”

This multi-faceted approach, as detailed by Sherrard-Smith, ensures that an idea is rigorously examined from various viewpoints before a conclusion is reached. He outlines the roles of these five simulated advisors:

• The Contrarian: Focuses on identifying the critical, potentially fatal, flaws in an idea.
• First Principles: Works to strip away assumptions and get to the foundational elements of a concept.
• The Executor: Questions the practical feasibility and implementability of the idea.
• The Expansionist: Seeks to uncover hidden upside potential that might be overlooked.
• The Outsider: Brings a fresh, context-free perspective to catch what internal or expert views might miss.

Implementing the LLM Council for Enhanced Feedback

Sherrard-Smith provides a clear, step-by-step guide for founders interested in adopting this more critical AI interaction model. He outlines the process for installing the LLM Council skill within Claude, making it accessible to users seeking deeper insights.

The steps involve downloading the skill, customizing it within Claude, uploading a specific file, and then activating the council. This practical advice underscores his commitment to helping founders leverage AI more effectively. As Dan Sherrard-Smith puts it:

“One Claude tells you you’re right. Five Claudes show you where you’re wrong.”

He concludes his post with a strong endorsement of the skill, signaling its value for founders aiming to avoid the pitfalls of AI-driven confirmation bias and instead foster genuine innovation through rigorous, multi-perspective feedback.