Lovable Advances AI App Building Beyond 'Vibes'

Lazar Jovanovic claims to have made over 37,000 edits in Lovable, highlighting the iterative nature of development. He argues that the industry is now ready to treat AI-powered development tools as production-grade systems, moving beyond the initial phase of 'vibe coding.' Jovanovic advocates for rethinking the use of Lovable beyond mere prototyping, emphasizing its potential for building robust applications.

He stated, "The one thing I learned 37,385 edits later, is that I should really stop calling this vibe coding. And that is mostly because I believe we're at the moment where we are way past vibes."

Ad-hoc prompting is described as inefficient and unpredictable, requiring developers to repeatedly input instructions and hope for consistent results. Lovable now allows users to write a 'skill' once, which enables AI agents to follow instructions consistently without the need for re-prompting.

These skills can be shared across different workspaces or imported directly from GitHub, promoting reusability and standardization in development. This shift from one-off prompts to persistent skills is central to making AI development more repeatable and scalable, supporting the creation of more complex applications.

Jovanovic noted, "Writing the same prompt from scratch and praying it lands well and praying that it actually works... If I want to build something repeatable and scalable, I should instead write a skill once and then the agent will know how to work its way around things every time that I want."

Lovable's skill-based architecture now enhances design flexibility, allowing users to invoke a 'redesign' skill that presents multiple design proposals. This new layer of flexibility is intended to bridge the gap between simple prototypes and real applications by offering creative options.

The platform also ensures that any subsequent design changes remain consistent with the established brand guidelines. This feature helps maintain visual integrity while still allowing for iterative design improvements, making the development process more adaptable.

Lovable introduces a visual editor that allows users to make UI changes by drawing directly on the screen. This method enables developers to visually annotate areas for modification, such as circling an element to change or crossing one out to remove it.

These visual inputs are then combined with text prompts to produce real-time results, significantly reducing the effort required to describe complex layout adjustments. The visual editing tool aims to simplify the iteration process and make it more intuitive for users.

Users have the flexibility to build these systems from scratch or base them on existing applications, providing a strong foundation for consistent design.

These custom design systems support various elements including screenshots, colors, typography, and reusable components. This feature prevents the AI from deviating from established brand guidelines, ensuring that all generated content aligns with the desired aesthetic and functional standards.

Jovanovic highlighted, "Lovable on all pro and business plans as well, supports bringing in and creating your own design system as a core source of truth for anything that you want to build."

This allows development decisions and constraints to be carried forward persistently, enhancing efficiency.

Users can define custom instructions that dictate the AI's behavior, ensuring consistent adherence to project-specific requirements. The AI confirms its use of this stored knowledge before proceeding with tasks, providing transparency and control.

The system includes persistent memory about the user and the ongoing build process, creating a more cohesive and intelligent development environment.

Lovable now includes integrated automatic security reviews for every build, addressing the complexities of managing secrets and APIs. The platform supports over 140 built-in connectors, streamlining integration with various services.

Custom APIs can be added securely without requiring users to paste sensitive keys directly into chats, enhancing security. Additionally, Lovable offers restricted collaboration features for private connectors, allowing controlled access to sensitive project components.

Lovable performs a basic security scan against code and dependencies with every code change, integrating security directly into the development workflow. Users are also encouraged to conduct deeper security scans as needed and can customize their security memory settings.

All security features and fixes, including these advanced scanning capabilities, are provided at no extra cost. This ensures that developers can maintain high security standards without incurring additional expenses.

Past development workflows often required developers to stitch together backend services across multiple dashboards and build processes, creating significant overhead. This often involved managing numerous separate accounts, such as 20-30 different Supabase accounts, due to project limits and isolated environments.

Lovable aims to simplify this by providing a more integrated approach to backend management. The goal is to reduce the fragmentation and complexity traditionally associated with configuring and maintaining the various components of a software's backend infrastructure.

Jovanovic illustrated the problem, stating, "The past used to look like stitching a back-end together across four dashboards and three bills."

Lovable has launched a beta feature for automated project monitoring, allowing errors to be flagged automatically even when users are not actively working. This system tracks code changes and error logs to identify issues proactively.

Users can configure monitoring schedules, choosing between daily scans or per-edit checks, based on their project's needs. The system sends email notifications for any discovered issues, enabling rapid response and reducing downtime.

Lovable has introduced enterprise-grade security features and compliance to meet the demands of large clients such as Uber, Nvidia, and T-Mobile. These enhancements include advanced PII detection and sensitive data scanning capabilities, ensuring robust data protection.

The platform now offers workspace-level security controls and usage limits, allowing organizations to manage access and resource consumption effectively. Lovable is also the first AI coding platform to achieve AICPA SOC 1 certification, underscoring its commitment to high security standards.

Jovanovic emphasized the importance of trust: "To be able to be enterprise-ready, you need to have trust built in from day one."

Lovable now provides an all-in-one suite for post-deployment management, enabling users to set up domains, payments, emails, and SEO from a single platform. The platform offers integrated payment support for Stripe and Paddle, simplifying transaction processing.

Users can benefit from a guided domain and email setup process that eliminates the need for manual API key management. Built-in SEO tools, powered by SEMrush integration, help optimize applications for search engines, while native analytics replace the need for third-party tracking pixels.

Jovanovic stated, "With Lovable, you can set up domains, payments, emails, SEO all in one spot."

Lovable has launched an interactive showcase at waypastvibes.lovable.app, where users can experience the presentation site directly. This new website template allows users to build their own personal versions of the presentation site.

The platform supports one-click deployment, making it simple for users to clone and customize the demo experience. This initiative aims to provide a practical example of Lovable's capabilities and encourage direct user engagement.

AI-assisted development still requires careful planning and specification, especially for complex projects, due to token limits that constrain the amount of work a single prompt can handle. Building complex applications, such as a Stripe clone, necessitates thorough planning and multiple Product Requirements Documents (PRDs).

Users must learn to distinguish between situations where 'vibe coding'—relying on intuition—is appropriate and when systematic planning is essential. This understanding of when to apply different development approaches is currently identified as one of the biggest challenges in AI-driven coding.

Jovanovic summarized this challenge, stating, "That will require some planning, some specs. So, I think that is actually the biggest challenge in vibe coding today. Understanding when to vibe versus when not to."

However, the industry is now transitioning towards 'AI Engineering,' emphasizing greater control, established security protocols, and adherence to production standards.

This shift is evidenced by the inclusion of security controls, security reviews, and distinct deployment processes for production versus development environments. By 2027, it is anticipated that AI agents will perform continuous autonomous testing and monitoring, further solidifying the move towards a more structured and automated engineering approach.

Lovable now allows users to trigger automated end-to-end tests using Playwright, a powerful browser automation library, to evaluate production readiness. The platform simulates various user actions, such as signing up and browsing, to comprehensively test application functionality.

This capability supports rapid iteration, enabling developers to identify and fix production bugs within minutes. The integration of robust end-to-end testing reduces the reliance on rigid upfront testing, allowing for more agile development cycles.