Vibe coding: the future of programming may be more spoken than written

Artificial Intelligence (AI) keeps advancing, and one of the most recent proofs of this is “vibe coding”. This concept represents a true transformation in the way developers and professionals from other fields interact with code: it is now possible to build software projects at the pace of a conversation, using natural language.

Vibe coding

The term exploded in February 2025, thanks to a post by Andrej Karpathy, a leading figure in the technology industry who has worked at Tesla and OpenAI. In the post, he described this new way of programming, in which text or voice commands can be used to write code, reducing the need for manual, repetitive typing.

The idea behind this new approach is to make AI such an efficient partner that developers can focus their attention on the outcome, instead of being tied down to writing the code. It is a more intuitive way of programming, one that speeds up the process and frees professionals to focus on error analysis and strategic decision-making.

For those outside the technical field, vibe coding also works as an assistant. Users can bring ideas to life without depending exclusively on feedback from the technology team, for example.

How it works in practice

In vibe coding, the main “language” used is natural language, such as English or Portuguese. The user describes what they need, and AIs like ChatGPT, Gemini, Claude, or DeepSeek turn these instructions into code with high precision. Tools like Cursor, GitHub Copilot, and Windsurf already integrate these models into their development workflows, making the experience even more fluid and collaborative.

While using it, it is important to understand that the developer's role changes: they take on the function of “director” and “reviewer”, while AI acts as a programming partner. It generates real code in widely used programming languages, such as Python, JavaScript, among others.

Because of this, those with little familiarity with coding should be careful not to confuse this approach with no-code and low-code. Unlike vibe coding, these concepts prioritize visual interfaces, precisely to minimize direct contact with programming.

And what are the next steps?

Given the speed at which AIs are evolving, vibe coding has the potential to be more than just “hype”. Even so, its impact has two sides.

On one side, the innovation can be an entry point for beginners, breaking down the intimidating barrier of syntax and enabling quick results. On the other, there is the risk of excessive dependence on AI, which can undermine mastery of programming fundamentals, essential for building complex applications and optimizing performance, in addition to potentially causing security issues, should the developer not know how to find them.

For now, vibe coding has been more recommended for creating prototypes, since developing a final product may demand more rigor, structure, and advanced programming skills.

Because of this, it is essential to adopt the tool with a strategic mindset: learning from it, without giving up continuous study and mindful practice. Using AI as an ally, without relying on it entirely, is the way to gain efficiency without losing technical depth.

Furthermore, it will become increasingly important to follow the debates on quality, security, and governance of generated code. These are topics that will certainly gain momentum in the coming years. And you, which approach do you prefer to use in your coding?