AI-powered logomaker built in 10 days challenges traditional design

The rise of “vibe coding” demonstrates how large language models have transformed software development, allowing even beginners to create functional applications with minimal coding knowledge. An experienced developer’s experiment with building a logo design tool entirely through AI-generated code reveals both the impressive capabilities and limitations of current LLMs when tasked with creating complete applications without human intervention.

The big picture: Developer Johnny Dunn built “Logomaker,” a functional logo design application, by exclusively using AI-generated code from tools like GPT-4o, Gemini Pro 2.5, Claude 3.7, and GitHub Copilot, without writing or editing any code himself.

• The experiment, completed in just ten days, demonstrates how developers can now build working software by simply describing requirements to LLMs and iteratively refining the AI-generated code.
• This approach, dubbed “vibe coding,” allows even coding novices to create functional applications by describing problems, testing solutions suggested by LLMs, and repeating until achieving usability.

Why this matters: The rise of LLM-assisted development is dramatically lowering barriers to entry in software creation while transforming how experienced developers work.

• Beginner coders can now produce user interfaces and experiences in days that would have taken exponentially longer just 3-4 years ago, simply by communicating with AI assistants.
• Experienced developers are similarly adopting workflows where they spend less time writing code and more time directing AI tools, fundamentally changing the nature of programming work.

Development approach: Dunn served primarily as a guide for the LLMs, allowing them to handle all code creation and refinement decisions.

• The developer deliberately avoided writing or editing any code personally, instead relying entirely on AI suggestions while applying his technical expertise to direct the process.
• Even refactoring decisions were left to the AI systems’ discretion, with Dunn stepping in only to provide guidance on the overall direction.

Key insights: The experiment revealed both impressive capabilities and significant limitations in current LLM-based development workflows.

• The resulting “Logomaker” application works as intended, demonstrating that AI systems can successfully generate functional web applications with minimal human coding intervention.
• The project’s code base reflects a “Frankenstein’s monster-esque creation” in JavaScript, highlighting how AI-generated code may lack the refinement and optimization of human-written alternatives.

The bottom line: “Vibe coding” represents a transformative approach to software development that will likely continue evolving as LLM capabilities improve, potentially reshaping skill requirements and workflows throughout the industry.