Vibe Coding: Can AI Really Give Us Free Photoshop and Excel?

What if AI could help recreate expensive software without the subscription fees. While the idea sounds appealing, creating a working alternative is not the same as replicating the full capabilities of established software. Understanding what vibe coding can achieve, and where it falls short, is essential before treating these claims as a practical solution.

What Is Vibe Coding?

Vibe coding is a way of developing software by describing what an application should do and letting an AI tool generate the code. The term was introduced in February 2025 by computer scientist Andrej Karpathy. Rather than writing every line manually, developers use natural language prompts to guide AI systems through the process. IBM describes this approach as a shift towards intent-driven software development, where users communicate their requirements and AI generates the underlying code.

However, generating software is only part of the process. The output still needs to be tested, reviewed and refined to ensure it works as intended. 

The Project Behind the Headlines

Brandon Thomas, an Atlanta-based software engineer with around 15 years of experience, has attracted attention for developing free, open-source alternatives to popular software. His project, ArtCraft, includes applications designed to replicate some of the features of Adobe tools such as Photoshop, Illustrator, and Premiere Pro. According to Al Jazeera’s report, Thomas developed the applications using Anthropic’s Claude Opus 5.5 and the Rust programming language. Frustrated with recurring subscription costs and cancellation fees, he set out to create alternatives that users could access without paying for expensive software licences.

He later introduced WordCraft, GridCraft, and DeckCraft, offering free alternatives to Microsoft Word, Excel, and PowerPoint. The projects are available as open-source software on GitHub. Thomas has also hinted at expanding the initiative to include alternatives to other subscription-based tools, as reported by Tom’s Hardware.

Challenges in AI-Generated Code

AI-generated code may work as intended but still contain security flaws. A 2025 Veracode report tested more than 100 AI models and found security vulnerabilities in 45% of the generated code samples. 

Thomas describes his applications as ‘clean-room’ rebuilds, developed using public file-format specifications rather than Adobe’s original code. Questions also remain about the extent of AI’s contribution and whether the development process meets legal standards.

Copyright presents another challenge. A 2025 US Copyright Office report concluded that AI-generated material is not automatically protected by copyright. Human contributions may qualify for protection when they involve sufficient creative input. This raises questions about ownership and licensing for AI-generated open-source software.

Is It a Win for Consumers?

Not quite yet. Free alternatives could make expensive software more accessible, but unfinished features, security concerns, and uncertain legal protections remain significant obstacles. There is also no clear evidence yet of how these projects might affect software industry jobs. Thomas’s experience as a software engineer highlights another important point: AI can accelerate development, but technical expertise is still needed to guide, test, and improve its output.

How to Use AI-Built Software Responsibly

• Treat AI-built clones as experiments: Do not rely on them as replacements for established software, especially for important tasks.

• Protect sensitive information: Avoid opening confidential files in applications that have not undergone independent security reviews.

• Check the project first: Review its licence, development activity, and available security assessments before installing or using it.

• Choose reliable tools for critical work: Established software, including trusted open-source alternatives, remains a safer option for tasks that require consistency and security.

In essence, ArtCraft demonstrates how AI can make software development more accessible, allowing individuals to create alternatives to expensive applications. However, their long-term success will depend on whether they can meet the security, legal, and reliability standards expected of established software.