Summary & Key Takeaways ​

• Thomas Ptacek suggests a strict rule: never use a single word an LLM suggests for writing.
• LLMs are best utilized for fact-checking, spelling, grammar, and as a thesaurus.
• The "weird smell" of LLM-generated text is a good indicator to avoid it.
• Adopting this principle helps maintain writing discipline.
• The article provides insights into building a personal LLM copyediting tool.

Our Commentary ​

"You may not use a single word an LLM suggests to you." I love this rule. It's so simple, so absolute, and it immediately resonates with that "weird smell" we all recognize in AI-generated text. We've been grappling with how to integrate LLMs without losing our voice, and this feels like a solid, actionable boundary. It's not about rejecting the tech, but about defining its role precisely. I'm definitely going to try adopting this for my own writing process.