Mastering yourself so AI can master your voice
Ruben Hassid explores the realization that human creativity and unique personal voice are ultimately a collection of patterns transferable to an AI via a simple text file. By shifting the focus from instruction to the “logic of refusal,” authors can successfully duplicate their distinct taste within LLMs. This strategic self-interrogation allows AI to move beyond generic averages to inhabit a specific human persona.
Points clés
- Ruben Hassid realized his “unique” writing voice could be captured entirely within a single .md text file.
- The author identifies that LLMs default to “statistical averages” and “slop” when they lack specific human context.
- Taste is defined more by what a creator rejects (refusals) than what they prefer.
- Hassid recommends a specific 100-question “DNA interview” to extract personal mechanics, beliefs, and “Aesthetic Crimes.”
- Claude (specifically Opus-4.5) is highlighted as the preferred model for following these complex identity instructions as of January 2026.
- The resulting markdown file makes a persona portable across different platforms like Claude, ChatGPT, Gemini, and Grok.
- “Mastering yourself” is presented as a prerequisite for AI to effectively duplicate your style.
- Ruben Hassid’s newsletter reaches over 238,000 subscribers, focusing on mastering AI without technical skills.
- The article itself was co-authored by AI using Hassid’s own “voice profile” text file.
- Paid subscribers to “How to AI” receive the author’s personal .md file as a template for their own duplication.
À retenir
So, it turns out you aren’t a tortured genius or a unique snowflake; you’re just a moderately complex script with a few pet peeves. If you want to stop sounding like a corporate chatbot, try looking in the mirror and actually deciding what you hate for once. It’s much easier to let a machine pretend to be you when you’ve finally admitted that your “soul” is mostly just a list of words you’re too pretentious to use. Happy cloning!
Sources
