Artificial intelligence researcher Andrej Karpathy, who coined the term "vibe coding", believes users don't need to spend time crafting the perfect AI prompt. Instead, they should simply explain what they want in their own words.
In a post on X, the former OpenAI co-founder and former Tesla AI director said he prefers using voice mode while interacting with Large Language Models (LLMs). Rather than typing carefully edited prompts, he speaks freely, even if his thoughts are unstructured or incomplete.
Karpathy stated that modern AI models are capable of organising scattered ideas into a clear response. He often tells the chatbot that his thoughts may be messy or contain mistakes, allowing the AI to focus on understanding the intent instead of interpreting every sentence literally.
He also argued that users often provide too little context. Based on his view, longer and more natural instructions can enable the AI to get the point better than shorter and more refined prompts. He says this amid the ongoing investment by AI companies in voice-driven assistance that enables more conversations.
In early 2025, Karpathy coined the term vibe coding to refer to a coding method that involves the use of AI to develop code from natural language prompts as opposed to writing everything by hand. He tells us to forget about perfect prompts and think about clear communication. That is what modern AI is all about.