Will the future of software development run on vibes?
1 min read
Summary
A new way of coding, termed “vibe coding”, has been coined by former OpenAI researcher Andrej Karpathy, describing a process of using large language models (LLMs) to create software without necessarily understanding how the code works.
Described as “surrendering to the flow” of the AI, users can communicate their intentions to the model in natural language, and the model will translate this into code.
As the approach relies on AI tools translating ideas into code, there are questions over whether it can produce code that is reliable for real-world applications.
Tools enabling this approach include Cursor Composer, GitHub Copilot and Replit Agent, making the process more accessible to those without programming expertise.
However, Karpathy admits this is ultimately the “ultimate lazy programmer experience”, with users able to be much less hands-on with the code, and feed errors back into the AI for it to resolve.
