Vibe Coding: The Creative Struggle Against AI Programming

Andrej Karpathy, the former OpenAI co-founder and AI visionary with a history at Tesla, recently revolutionized the programming lexicon with his concept of vibe coding (Karpathy, 2024). Now, you might be wondering: what exactly is vibe coding? It’s that moment when you throw caution to the wind, embrace the vibes, and forget that the code even exists (Karpathy, 2025). In more formal terms, vibe coding refers to the act of crafting a program primarily by prompting generative artificial intelligence (AI). Sounds easy, right? 😏

🚦 The Buzz: Is Vibe Coding the Future?

There's a tangible buzz in programming circles about vibe coding potentially becoming the new normal. Andrew Chen, a former Uber exec and now a venture capital aficionado, boldly asserts that vibe coding is set to disrupt the programming landscape dramatically. In his article titled “Vibe Coding: Some Thoughts and Predictions” (Chen, 2025), he forecasts that vibe coding will:

• Democratize software production
• Commodify software
• Normalize AI-augmented creation

While these predictions ignite excitement, they're also accompanied by a hefty dose of skepticism. 🤔 Let’s navigate through this with our critical goggles on.

❌ The Reality: Vibe Coding Misfires?

Sure, AI is a game-changer, but can it really unlock the creativity that makes software shine? Allow me to pose a counterpoint: AI lacks the human heart (because, obviously, it doesn't have one), and creativity along with problem-solving is inherently human.

📉 The Dangers of Vibe Coding

1. Democratization Doesn’t Equal Creativity
   Chen envisions a future where kids and everyday folks become software creators thanks to vibe coding. But the reality? The magic of crafting meaningful, innovative, and tailored software comes from deep understanding. Can you really expect a novice to craft masterpieces like Shakespeare wrote plays?

2. From Alive to Undead
   According to Paul Naur’s theory from 1985, programs aren’t merely codes; they are the living mental models of their creators. These ‘living programs’ with inherent theory can be a joy to modify and evolve. When AI produces outputs, they often resemble zombies—technologically functional but lacking in actual ‘life’ (Naur, 1985).

3. Model Collapse is Real
   When AI starts training on its own output without the necessary human intervention, it risks descending into a realm known as model collapse. This is when AI begins confusing previous errors for reality (Shumailov et al., 2024). Yikes! 🧟‍♂️

🎨 Programming is an Art

Let’s rally around the idea that programming is art, much like painting or sculpture. As an advocate, I believe that beautiful programming (which can be seen as a form of art) provides elegance and sophistication that AI simply cannot replicate. Daniel Knuth, an influential computer scientist, argues that the potential to write beautiful programs is what draws many into the coding world.

🎨 What Makes Programming Art?

• Evocation: Inspiration from external stimuli.
• Transcendence: Personal unique awareness of what to create from inspiration.
• Motivation: The innate urge to create.

AIs may capture evocation and motivation, but they critically lack transcendence. Thus, AI-generated programs must rely on learned patterns rather than creative thought, which dilutes their essence as genuine software ( Oleynick et al., 2014).

😒 AI or Human: Who’s the Better Coder?

Consider this: Bill Gates argues that coders, alongside specialists in biology and energy, will continue to thrive in a world dominated by AI (The Economic Times, 2025). When pitted against algorithms like ChatGPT (GPT-4), many studies indicate that human programmers still outperform their AI counterparts in complexity, logical structuring, and adherence to edge case handling (Azeem et al., 2025).

🏁 Final Thoughts: The Future of Coding

While vibe coding may create functional programs, they ultimately lack the artistry and driven intent of human programming. Each piece of software retains the imprint of its creator, fostered through human ingenuity and creativity. This distinction is pivotal and ultimately warrants a compelling argument: coding should not only be a mechanistic exercise within the confines of AI applications; it should be about maintaining the human touch in a dance with technology we need to celebrate. Thus, we must tread thoughtfully into the future—because if vibe coding becomes the norm, we risk losing our identity as programmers.

📚 References

• Azeem, N. S., Naveed, N. M. S., Sajid, N. M., & Ali, N. I. (2025). AI vs. Human Programmers: Complexity and Performance in Code Generation. VAWKUM Transactions on Computer Sciences, 13(1), 201–216. Link
• Bohacek, M., & Farid, H. (2023). Nepotistically trained Generative-AI models collapse. arXiv (Cornell University). Link
• Chen, A. (2025, March 10). Vibe coding, some thoughts and predictions. @andrewchen. Link
• Dunsin, D. (2025, March). Ethical and Security Implications of Using Generative AI in Software Development. ResearchGate. Link
• Naur, P. (1985). Programming as theory building. Microprocessing and Microprogramming, 15(5), 253–261. Link
• Oleynick, V. C., Thrash, T. M., LeFew, M. C., Moldovan, E. G., & Kieffaber, P. D. (2014). The scientific study of inspiration in the creative process: challenges and opportunities. Frontiers in Human Neuroscience, 8. Link