Ruben Hassid’s Method for Replicating Your ‘Brain’ in AI Assistants

In a recent LinkedIn post, Ruben Hassid explores a novel method for replicating an individual’s thought processes and knowledge base within AI large language models, specifically using Claude. Hassid frames this as a way to ‘duplicate your (entire) brain’ in an AI, enabling more consistent and personalized AI interactions.

Hassid outlines an 11-step process that involves a combination of downloading specific AI applications, configuring settings, and meticulously preparing personal knowledge files. The core of his strategy hinges on creating structured data files that the AI can reference, thereby imbuing it with the user’s unique context, preferences, and operational guidelines. As Ruben Hassid notes:

“Cowork = where your brain lives”

This ‘Cowork’ mode, accessible within the Claude desktop app, is presented as the central hub for managing these personal knowledge files. Hassid emphasizes the importance of voice input, suggesting that dictating context is faster and more ‘honest’ than typing, a sentiment he captures with the provocative statement, “Typing kills the truth. Dictate your context.”

The ‘Taste Interview’ and Knowledge Compression

A significant part of Hassid’s methodology involves a detailed ‘taste interview’ and subsequent compression of this information into manageable files. He directs users to a specific prompt designed to elicit 100 questions across seven categories, pushing for specificity to avoid vague answers.

Creating Foundational AI Knowledge Files

Following the interview, Hassid details the creation of two key Markdown (.md) files:

• Anti-AI-Writing-Style .md: This file is crucial for preventing the AI from defaulting to its generic output. Hassid advises users to list words and grammatical structures they wish to avoid, such as “delve” or negative parallelism. This ensures the AI’s responses align with the user’s preferred communication style.
• My-Company .md: This file acts as a ‘north star,’ containing the user’s core goals, focus areas, and immediate ‘hard no’s.’ Hassid recommends keeping this file updated quarterly and ensuring it exceeds 1,000 tokens to provide a robust directional guide for the AI.

Hassid argues that without these stylistic and contextual guardrails, Claude, or any other AI, will simply “write like Claude.”

Testing and Deployment Across Platforms

Before final deployment, Hassid suggests testing the setup in a blank chat. The prompt prompts users to generate a response that genuinely sounds like them. Once satisfied, the user is instructed to drop all three files (the ‘about-me,’ ‘anti-AI-writing-style,’ and ‘my-company’ files) into the ‘Brain’ folder within Claude’s Cowork mode. According to Ruben Hassid, “Claude reads them on every single turn,” ensuring consistent application of the user’s persona and knowledge.

The final step, ‘Port it everywhere,’ highlights the portability of this method. Hassid suggests that the same structured files can be uploaded to other AI platforms like ChatGPT, Grok, and Gemini, effectively allowing users to maintain a consistent ‘brain’ across different AI tools. He further notes the potential for handing these files to a team or ghostwriter, standardizing output and communication.

“You stop prompting your brain. Your files do it.”

Hassid acknowledges that this process might feel daunting, stating, “You’ll resist this. It feels reductive. It feels scary.” However, he counters this by suggesting that perceived complexity often stems from vagueness. By creating these explicit, structured files, users can overcome this ambiguity and achieve a more precise AI representation of themselves and their professional context.

For those interested in acquiring Hassid’s specific folder and template files, he directs them to subscribe to his free newsletter, where a welcoming email provides access to the Notion link containing the downloadable assets. This allows individuals to become the designated ‘AI guy’ within their organizations by leveraging Hassid’s structured approach.