Vibe Coding != Developing with AI Assistance
🚀 Vibe Coding != Developing with AI Assistance 🤖
On February 2nd, Andrej Karpathy, former Head of AI at Tesla and co-founder of OpenAI, posted on X about a new way of developing software he calls Vibe Coding, where AI generates code from natural language descriptions while the human guides and refines the outcome forgeting that the code even exists. However, he cautioned: “it’s not really coding.”
In line with this, Simon Willison — widely known as the creator of Django, one of the most used Python frameworks — argues that using AI to assist with coding is not Vibe Coding, for several reasons, including:
👨💻 Professional programming demands understanding, review, and quality.
🔐 Security and privacy are critical risks that must be considered.
⚠️ Vibe Coding is only appropriate for low-risk projects.
✅ It’s essential to review, test, and understand the generated code.
The debate around Vibe Coding is active, and like any new trend, it has both supporters and skeptics, advantages and challenges. Among the advantages, we find:
- Enables non-programmers to create functional software.
- Speeds up the development of prototypes and MVPs.
- Facilitates experimentation and innovation.
As for the challenges:
- Potential bugs and vulnerabilities in the generated code.
- Risk of relying on limited understanding of the code by users.
- Significant risks in production environments without proper code review.
📊 Regardless of terminology, the trend toward using AI to build software is undeniable. A survey conducted in early 2024 by Inbal Shani, GitHub’s Chief Product Officer, across 5,000 U.S. companies revealed:
- 92% of developers already use AI in their daily work.
- 70% believe AI tools give them a competitive advantage.
Additionally, according to EliteBrains, AI generated 41% of all code in 2024, amounting to 256 billion lines of code, indicating its widespread adoption.
Similarly, in a March 10th interview hosted by the Council on Foreign Relations, Dario Amodei, CEO of Anthropic, stated that within the next 3 to 6 months, AI will write 90% of the world’s code — and by next year, 100%. This prediction was made during the discussion at approximately minute 16:16.
These figures align with statements by Jared Friedman, Managing Partner at Y Combinator, who reported that in Q1 2025, 25% of startup founders backed by the fund said that 95% of their code was generated by AI. This insight was shared during a podcast episode of the Y Combinator Startup Podcast, in response to the question: "What percentage of code is being written by LLMs?" around minute 10:01.
Without a doubt, we are witnessing a new way of building software — a profound shift in the industry that requires knowing when to let the vibe flow… and when to take control.
Want to dive deeper? Check out the resources bellow! 🔍
Andrej Karpathy's Vibe Coding post 👉 Vibe Coding: A New Way of Building Software
GitHub survey 👉 Survey reveals AI's impact on the developer experience
Dario Amodei interview 👉 The Future of U.S. AI Leadership with CEO of Anthropic Dario Amodei
Y Combinator podcast 👉 Vibe Coding Is The Future
EliteBrains report 👉 AI-Generated Code Statistics 2025: Is Your Developer Job Safe?
