Yonathan Cohen Reveals Method for Merging AI Responses on Claude

In a recent LinkedIn post, Yonathan Cohen shares a unique approach to leveraging multiple AI models simultaneously, aiming to synthesize the best aspects of each into a superior final answer. Cohen details a “Skill” he has developed within Claude that allows users to prompt three different AI large language models – ChatGPT, Perplexity, and Claude itself – in parallel, with a mechanism to consolidate their outputs.

Synthesizing AI Strengths

Cohen’s method addresses a common challenge with AI tools: while powerful individually, each model has its own strengths and weaknesses. By running queries through multiple AIs at once, his “Skill” aims to capture the most effective parts of each response while discarding less useful information. As Yonathan Cohen explains the core function of his technique:

I drop 1 prompt in Claude. ChatGPT, Perplexity & Claude answer in parallel. A final card keeps what’s genuinely good from all three.

This parallel processing and consolidation is designed to produce a more robust and comprehensive answer than any single AI could provide. Cohen highlights the benefits of this integrated approach, stating:

Keeps the strong points from each. Drops the weak ones. Picks a side when they disagree.

The ultimate goal, according to Cohen, is to achieve a higher quality of output. “1 answer better than any single AI gave you,” he asserts, underscoring the perceived advantage of his multi-AI strategy.

Setting Up the AI Skill

For interested users, Yonathan Cohen outlines a straightforward setup process within the Claude platform. The steps involve navigating to Claude’s settings, accessing the “Connectors” section, and then adding the specific “Skill” he has created. Once integrated, users can begin employing the multi-AI prompting technique.

A Competitive Approach to AI Answers

Cohen contrasts his method with the more conventional use of AI, where individuals typically rely on a single tool for information. He frames his approach as a competitive dynamic for AI models. “Most people trust 1 AI for the answer. I make 3 compete and keep the best,” Cohen writes, positioning his technique as a way to harness the collective intelligence of different AI systems.

To encourage engagement and share the skill, Cohen prompts readers to comment “SKILL” on his LinkedIn post, indicating a desire to disseminate this innovative AI workflow.