Prioritising speed and experimentation
Vibe coding is a modern approach to software development where features and systems are described using natural language prompts and an AI system builds the code to bring these ideas to life. Unlike traditional software development, which tends to be built methodically and logically, Vibe coding prioritises speed and experimentation. In addition, as the AI system is writing the code, Vibe coding can be an effective way for non-developers to quickly build small, low-risk tools to improve efficiency and productivity.
How does Vibe coding work?
Prompt the AI: The user provides instructions using natural language prompts (known as 'vibes') that describe the system's functionality.
The AI generates the code: The LLM generates the code based on the natural language prompt.
Ask the AI to refine: Instead of directly reviewing the code, the user interacts with the AI to make amends, provide feedback or solve bugs.
Focus on the big picture: The quick, iterative approach encourages experimentation with big changes rather than the precise engineering of individual blocks of code.
Where is Vibe coding useful?
Vibe coding is brilliant for small, low-stakes projects in scenarios where the value is in productivity and speed rather than perfection. It shouldn't be used for business-critical systems or anything involving payments or personal data. Instead, it's well-suited to side projects, throwaway tools and experimental applications.
It can be a way for beginners to get started with coding or to help experienced developers build systems in languages less familiar to them. At its core is a paradigm shift where you no longer need to learn programming languages and syntax to develop software. Instead of writing code, the AI handles implementing your ideas so you can concentrate on creating software that solves real problems and adds value to your business.
Vibe coding will unleash a wave of creativity
AI is rapidly changing software development and we've quickly gone from systems that help correct syntax errors and autocomplete code to ones where AI agents can build working applications themselves. This is not only empowering developers to become much more productive, it's also democratising software development.
Mario Rodriguez, GitHub's Senior VP of Product, envisions AI tools like Copilot transforming software development. He anticipates a significant shift where coding not only becomes more accessible but also more intuitive and experimental:
"It's gonna feel a lot more like how kids play. It's like you create something you play with and you're like, Nope. Then you instruct it again… It's going to be real-time development."
Professional developers can also benefit from this wave of creativity. With faster prototyping and testing of new ideas, feedback from users can come much earlier, reducing delays and frustrating code rewrites.
Vibe coding will appeal to managers (and clients)
With the help of AI there are going to be a lot more people that can build software, but the people building the tools won’t necessarily understand how the code works. However, this has always been the case;
Manager (or client) describes a new feature to a developer using natural language
Developer implements the feature using code
Manager tests the app by using it rather than reading (or understanding) the code
Manager reports bugs to the developer using natural language
Developer makes changes to the code to fix the bugs
Manager tests the app by using it rather than reading the code (rinse & repeat)
Of course, you can have managers (like myself) who understand the code and how it works. However, more often you will have a situation where a manager has no idea how the code works and is completely reliant on the developer (or the AI model) to keep it running smoothly.
When you are working with professional developers, either in house or through an agency, it's that relationship that helps managers and clients to manage risk. You don’t need to understand the code when you have employed a developer who does. You can own the software and the benefits it brings but have the code managed on your behalf.
‘But what if they leave?’ Well, you can put another system in place to manage that risk such as hiring another developer, improving documentation or bringing an agency on board for continuity (we do this a lot).
Managing risk in software development
There are existing and well established ways to manage risks within software development. A lot of the risks associated with Vibe coding are very similar to the challenges faced by developers;
Without clear guidelines or a systematic structure, the code may become unorganised, making it difficult to maintain and understand.
Where code lacks proper documentation, it is hard for others (or even the original developer) to follow the logic and purpose of the code.
As projects grow, code that previously worked well can need considerable rewriting in order to scale, and code structure then becomes more important.
Developer’s skill levels can vary but code needs to follow standard practices to work securely and to allow other people to understand it.
Effective code needs to be reusable and should be designed with a larger plan in mind.
I’m not saying that AI generated code doesn’t have this, in fact some of the more modern agentic AI setups can do this very well. The problems start to happen when the operator of the AI agent doesn’t understand good coding practices and can provide instructions that on the face of it look fine but in the code create a mess that not even the best, professional (human) developer could possibly understand.
The positives still far outweigh the negatives
There may be risks associated with Vibe coding but the positives around AI assisted development still outweigh the negatives. With the help of AI systems, more and more people can build software themselves. That has the potential to provide enormous efficiency benefits to businesses everywhere.
AI tools are already improving developer productivity and this is only going to increase. Clients are going to be able to expect more for their money as the cost of development decreases. When an agency previously might have been able to show a few concept screens to a client, they could soon be presenting working demos. Old, vulnerable code that is too expensive to change can finally be removed and replaced with the help of AI.
Here at Cursor, we are investing in AI research and evaluating new tools and technologies as they become available. As professional developers ourselves we aren’t using Vibe coding on any client projects - but it is important that we know about it. AI is changing the face of our industry and just like the early days of the Internet it's a very exciting time to be involved in software development.
