New Way to Make Apps Uses Simple Words, But Good Instructions Are Key

A new way to build apps lets people use everyday words instead of code. This is called 'vibe coding.' It's helpful, but the quality of the app depends a lot on how clearly you tell the AI what to do. Good instructions lead to better apps.
By Newsroom | Technology, Artificial Intelligence |

Recent developments in AI suggest a shift towards "vibe coding," a method where users describe app features using natural language instead of traditional code. This approach allows individuals without deep programming knowledge to create functional applications. While the allure of rapid, intuitive development is strong, the actual effectiveness and quality of these AI-generated applications appear to be heavily dependent on the precision and detail of the user's prompts. Early explorations indicate that while the barrier to entry is significantly lowered, achieving robust and predictable results requires a more sophisticated understanding of how to instruct the AI, leading to a debate on whether "vibe coding" truly replaces, or merely augments, the need for skillful prompt engineering.

I Vibe Coded an App With 3 Popular Chatbots. The Real Winner Is a Good Prompt - 1

Context

The practice of "vibe coding" has emerged as a noteworthy trend in application development. This technique allows individuals to create applications by describing their desired features using everyday language, rather than writing code in a programming language. The process typically involves interacting with AI chatbots that translate these natural language instructions into functional code.

Read More: UK Government Wants New Online Safety Rules for Kids

I Vibe Coded an App With 3 Popular Chatbots. The Real Winner Is a Good Prompt - 2

  • Timeline of Vibe Coding:

  • Early 2025: The concept of "vibe coding" began to gain traction, with early examples appearing on platforms like Zapier, showcasing how AI could assist non-developers in building simple apps and websites. Applications like "Lovable" were noted for their use in this domain.

  • Mid-2025: The underlying technology and user experience of AI chatbots became more accessible. Discussions and guides emerged on how to create chatbots from scratch and integrate them into the vibe coding process, detailing the use of various AI providers like Groq, OpenAI, and Claude.

  • Late 2025: The focus began to shift towards practical application and quality. Step-by-step guides were published on how to build "vibe coded" apps that are "actually worth shipping," emphasizing the importance of a structured approach and the right tools, such as Replit, for development. The discussion also started to include the nuance of prompt crafting, with practical examples and tips appearing.

  • Ongoing: The debate regarding the efficacy of "vibe coding" versus traditional prompt engineering continues, with a growing recognition that the quality of the prompt directly correlates with the quality of the output.

  • Key Actors & Tools:

  • Users/Non-Developers: Individuals seeking to build applications without extensive coding knowledge.

  • AI Chatbots/Models:

  • Lovable: An AI coding app for front-end development.

  • Claude: Developed by Anthropic, used for various AI tasks.

  • Gemini: Google's AI model, accessible via Google AI Studio and Vertex AI.

  • OpenAI Models (GPT): Including GPT-4o and GPT-3.5-turbo.

  • Groq: A platform that allows for dynamic model selection.

  • Development Platforms:

  • Zapier Agents: Allows building AI agents by connecting work apps.

  • Replit: An integrated development environment that supports AI-assisted coding and database scaffolding.

  • Gemini CLI: A command-line interface for interacting with Gemini.

Read More: AI Systems Can Now Work Together Better

Evidence

The efficacy of "vibe coding" is directly tied to the user's ability to formulate precise and comprehensive instructions for the AI.

I Vibe Coded an App With 3 Popular Chatbots. The Real Winner Is a Good Prompt - 3

  • "AI chatbots take the prompts you write and pass them to their AI model." (Article 5)

  • "Unless you feed your AI with a clear design system, your app may function unpredictably." (Article 7)

  • "Not every prompt gives you clean results. You get predictable, structured output that respects your coding conventions. Each time you refine the prompt, you improve both clarity and results." (Article 4)

  • "Vibe coding is a programming technique where you describe what you want to create using natural language and have an AI tool build the app for you." (Article 7)

Vibe Coding in Practice: Tools and Examples

Vibe coding is facilitated by various AI tools and platforms, enabling users to translate natural language into code. The choice of AI provider and development environment can influence the outcome.

Read More: Bill Gates Likes Andhra Pradesh's Use of Technology

I Vibe Coded an App With 3 Popular Chatbots. The Real Winner Is a Good Prompt - 4

  • AI Providers: Users can select from a range of AI backends, including Groq, OpenAI, Claude, and Gemini. Each provider utilizes different models (e.g., Claude-3-Sonnet, GPT-4o, Gemini 2.0). (Article 2)

  • Development Environments: Platforms like Zapier Agents allow for the creation of AI agents that can connect to other work applications. Replit offers options for generating full application builds, including database structures, or starting with a design prototype. (Article 1, Article 3)

  • Application Examples:

  • A portfolio website AI coding app used Lovable. (Article 1)

  • A Plywood Cutting Visualizer AI coding app utilized Claude. (Article 1)

  • An admin tool for Supabase was built using Lovable, v0, and Cursor. (Article 1)

The Critical Role of Prompt Engineering

While "vibe coding" simplifies the initiation of app development, the quality and predictability of the generated applications hinge on the user's prompt engineering skills. A well-crafted prompt leads to better, more usable code.

Read More: OpenAI Executive Fired After Safety Concerns and Discrimination Claim

  • Prompt Refinement: Iteratively refining prompts is crucial for improving both the clarity of the instructions and the quality of the resulting code. (Article 4)

  • Specificity: Users are advised to be specific in their prompts to avoid unpredictable application behavior. (Article 7)

  • Predictable Output: Effective prompts can yield predictable, structured output that adheres to coding conventions. (Article 4)

  • Prompt vs. Coding: Some suggest that focusing on advanced prompt engineering might be more beneficial than the broader concept of "vibe coding" itself, implying a need for deeper skill in instruction. (Article 8 - Note: This source had low priority and a brief summary)

Evaluating Vibe Coded Application Quality

Beyond initial creation, the maintainability and reliability of applications built through vibe coding are significant considerations. Ensuring these qualities requires attention to specific coding traits and error handling.

Read More: Microsoft AI Boss Says Many UK Office Jobs Could Be Done by AI Soon

  • Key Traits: Desirable attributes for coded applications include:

  • Usability

  • Maintainability

  • Security

  • Reliability

  • Scalability

  • Availability

  • Flexibility (Article 6)

  • Error Handling: Developers are encouraged to identify critical code paths (such as database calls, API requests, and file I/O) to ensure robust error handling. (Article 6)

  • Design System: A clear design system provided to the AI is essential for predictable application function. (Article 7)

Conclusion

The emergence of "vibe coding" democratizes application development by enabling individuals to create software using natural language. Tools and platforms are readily available to support this process, with various AI models and development environments offering different capabilities. However, the practical success and robustness of these AI-generated applications are not guaranteed by the method alone. Evidence strongly suggests that the precision, clarity, and detail of the user's prompts are paramount. Without skillful prompt engineering, applications may exhibit unpredictable behavior and lack essential qualities like maintainability and reliability. The ongoing discussion points to a future where mastering prompt design will be a key determinant in the effectiveness of "vibe coded" projects, potentially blurring the lines between intuitive creation and disciplined engineering.

Read More: New Chip Works Like a Brain to Help Robots See Faster

Sources Used:

Frequently Asked Questions

Q: What is vibe coding?
Vibe coding is a new way to make apps. You use normal words to tell the computer what you want, and AI builds the app for you.
Q: Do I need to know how to code?
No, you don't need to know traditional coding. You just need to explain your ideas clearly in words.
Q: Does vibe coding always make good apps?
The apps work best when the instructions, or 'prompts,' are very clear and detailed. Simple or unclear instructions can lead to apps that don't work as expected.
Q: What is prompt engineering?
Prompt engineering means writing very good and specific instructions for the AI. This helps the AI create the best possible app for you.
Q: What tools can I use for vibe coding?
You can use AI tools like Claude, Gemini, or OpenAI models, and platforms like Zapier or Replit to help build your apps.