Lovable, the AI app builder that reached unicorn status last summer, is facing intensifying competition from rivals who are taking a different strategic approach: building their own AI models instead of relying on third-party systems. While Lovable has achieved remarkable growth, reaching $500 million in annual recurring revenue (ARR) earlier this month, the broader vibe-coding landscape is shifting as platforms seek to differentiate themselves and reduce costs.

Why Are AI App Builders Building Their Own Models?

Base44, the vibe-coding platform acquired by Wix for $80 million just one year after its founding, recently began rolling out its own custom large language model (LLM), called Base1. An LLM is an AI system trained on vast amounts of text data to understand and generate human language. The move reflects a broader industry concern: whether businesses built entirely on top of someone else's AI models can remain defensible long term.

According to Maor Shlomo, Base44's founder, owning the model provides significant advantages. "Training and owning the model as part of our entire stack allows us a lot more optimizations on latency, cost, and efficiency," Shlomo explained. Base44 developed Base1 using a dataset generated from tens of millions of real user interactions on its platform, giving it specialized knowledge tailored to app creation workflows.

The cost pressure is real. As inference costs (the expense of running AI models) have become a meaningful part of the business equation, enterprise customers are increasingly demanding optimization. Jonathan Userovici, a general partner at venture capital firm Headline, noted that companies are setting up entire infrastructure systems to select the right models for different tasks, ensuring costs don't skyrocket while maintaining performance.

What Makes a Vibe-Coding Platform Defensible?

Defensibility in the AI startup world comes down to three key ingredients: data, distribution, and tech stack. Base44 is betting that vertical integration, meaning ownership of all three elements, will cement its position as the only vibe-coding platform that controls its distribution, data, and infrastructure simultaneously.

However, the competitive landscape extends beyond other vibe-coding startups. Frontier AI labs, the companies building the most advanced general-purpose models, are moving into Base44's territory. Cursor and xAI (Grok's parent company) both belong to SpaceX, and Claude Code, made by Anthropic, has become a vibe-coding player in its own right. These foundational AI providers have access to data and feedback loops that help them improve models for app creation.

Shlomo believes specialization gives Base44 an edge. "Models are progressing, but they'll stay very general in what they can do," he predicted. Yet Userovici cautioned against underestimating frontier models, citing the example of Harvey, a legal tech startup that abandoned plans to train its own model and instead focused on optimizing how it uses existing systems.

Shlomo

How Are Designers and Developers Using Vibe-Coding Tools?

Vibe coding, a term coined in early 2025, represents a fundamental shift in how software gets built. Instead of writing code line by line, developers and designers describe what they want in natural language, and AI generates the entire application. Tools like Lovable allow users to import designs from Figma, a popular design tool, and generate corresponding React components with styling, then refine them through conversation.

For designers especially, vibe coding opens new possibilities. Rather than depending entirely on developers to bring ideas to life, designers can now control more of the creation process. Brandon Groce, a designer and educator, demonstrates this approach by using ChatGPT and Base44 together to create minimum viable products, proving that designs work before expensive development begins.

However, vibe coding has clear limitations. It works well for specific project types, such as business websites, portfolio sites, and simple e-commerce stores. It is generally unsuitable for complex, enterprise-grade applications requiring high security, strict maintenance standards, or rigorous testing protocols. Additionally, prompts need technical specificity; telling AI to "make it user-friendly" produces poor results, but specifying "use 16px minimum font size" and "ensure 44px touch targets" yields better outcomes.

Steps to Get Started With AI-Assisted Development

• Start with quick fixes: Approach your development team about contributing minor bug fixes or UI adjustments. This builds credibility without requiring full vibe-coding expertise and frees up developer time for more complex work.
• Build micro-interactions for testing: Generate working versions of hover states, animations, or loading states to validate interactions with users before handing specifications to engineering teams.
• Create small features: Add simple filters to dashboards, generate CSV export functions, or create basic settings panels. These targeted additions let you practice AI coding without the complexity of complete applications.

The key is recognizing that AI generates code quickly, but human judgment ensures it serves real people solving real problems. Designers and developers who combine AI's capabilities with an unwavering focus on user needs, accessibility, and security will thrive in this new landscape.

Why Are Users Looking Beyond Lovable?

While Lovable remains popular for transforming prompts into visually appealing apps, users are increasingly seeking alternatives to meet production-level demands. The reasons include limited customization options, concerns about vendor lock-in, and the desire for full code ownership to ensure long-term maintainability.

Production software requires far more than fast UI generation. Scalability, backend flexibility, security controls, and collaboration workflows are essential. As projects evolve, users often want deeper control over databases, APIs, authentication systems, and deployment environments, capabilities that prompt-only tools may not provide.

Pricing is also a concern for growing teams. AI-powered tools can become expensive as usage scales, especially for projects with frequent iterations, multiple collaborators, or large codebases. Additionally, developers are increasingly prioritizing production reliability; tools designed primarily for rapid prototyping may struggle with larger applications, advanced debugging, version control workflows, or enterprise-grade security requirements.

Base44's growth trajectory suggests the market is expanding rapidly. The platform passed $150 million in ARR in May, just two months after crossing $100 million in ARR. While this is still less than Lovable's $500 million in ARR, Base44's momentum and strategic focus on cost optimization and specialization indicate that the vibe-coding market is becoming increasingly competitive.

The broader lesson is clear: as AI-powered development tools mature, differentiation will come not just from speed and ease of use, but from ownership of data, control of infrastructure, and the ability to optimize costs for enterprise customers. Lovable's dominance in the space is no longer assured, and the next chapter of vibe-coding will be written by platforms that can balance rapid development with production-grade reliability and defensibility.