March 2025 Summaries

Mindaugas, founder of Coho, transitioned from recruitment and community-building into product development using Lovable. Without prior coding experience, he built a job search guidance tool that gained paying users within days, demonstrating that the future of education isn't in courses but in building real things. He leveraged AI-driven tools to accelerate learning, solve real problems, and overcome traditional constraints. Mindaugas' journey shows designers have a unique advantage in AI development due to their expertise in user experience and visual language. By embracing no-code and AI-assisted development, individuals can turn ideas into reality without needing extensive coding knowledge, paving the way for a new era of innovation.

Lovable experienced two availability incidents on Friday, affecting only its main website while published user applications remained unaffected. From 9:20 to 10:10 UTC, the Lovable site suffered from slow loading times due to an outage at Posthog, a third-party vendor responsible for running experiments and feature roll-outs, which delayed the site's startup by up to 20 seconds. Later, from 12:00 to 12:45 UTC, an incident with their hosting vendor, Netlify, rendered the website completely inaccessible until service was restored. Both issues originated from upstream vendor problems, prompting Lovable to expedite plans for local resolution of feature flags and evaluate alternative hosting options to enhance resilience against similar disruptions in the future.

Lovable has introduced several updates, including History 2.0, which enhances versioning with improved navigation, bookmarking, and restoration features, making it easier to track changes and undo mistakes. Dev Mode now allows users to view and edit code within the Lovable Editor without needing a GitHub account connection, with code editing available as an experimental feature for paid users. Various small fixes have been implemented, such as resolving server errors and syncing Labs Features across devices. Additionally, Lovable will host a livestream event focusing on app-building with Clerk for advanced user management, featuring a live demo on integrating user authentication with Supabase. The event is scheduled for March 13th at 19:00 CET, and participants can register online.

When upgrading to Claude 3.7 Sonnet, a performance bottleneck emerged in the Diff Viewer, causing browser freezes when many files were modified due to the simultaneous loading of multiple CodeMirror instances. Each instance took 20-50ms to load, creating long blocking tasks, especially when 15 or more editors loaded concurrently. This was further exacerbated by forced reflows during CodeMirror's initialization, a known issue. Several potential solutions were considered, including collapsing files by default or using an intersection observer, but these either required manual intervention or introduced complexity. Ultimately, a time slicing approach was adopted, which involved batch loading two files at a time, using placeholders for others, and implementing small delays between batches to maintain browser responsiveness. This strategy transformed a single long task into multiple shorter ones, improving both performance and user experience by allowing content to appear progressively and naturally, thus preventing the jarring effect of all diffs expanding simultaneously.

Lovable, initially a front-end text-to-app platform, integrated with Supabase to overcome backend limitations such as the need for databases, authentication, and code execution. Supabase was chosen for its open-source nature, Postgres foundation, and user-friendly interface, among other features. However, deploying Supabase edge functions required workarounds due to the necessity of using a CLI and Docker, which Lovable managed by orchestrating Supabase integrations seamlessly. This led to significant growth for Lovable, increasing its ARR from under $500K to $20M. The integration highlighted the importance of context in AI interactions, prompting the development of the Model Context Protocol (MCP), which aims to standardize how SaaS platforms expose resources to LLMs. MCP promises to streamline AI-native integrations by allowing SaaS companies to provide structured data, ultimately creating faster integrations and more powerful AI applications, potentially marking a significant advancement in AI-driven development.

In a draft article, the development of the Visual Edits feature for Lovable, an AI-driven full-stack app-building platform, is detailed as a significant technical achievement aimed at giving developers and designers more precise control over their applications. This feature, akin to a Figma-like editor, allows users to visually edit and save frontend changes directly in their apps, addressing the traditional frustrations of UI development cycles. By integrating an in-browser coding layer and employing techniques such as stable JSX tagging, client-side AST manipulation, and seamless hot module replacement, Lovable ensures rapid iteration and high-quality code generation. The implementation of Visual Edits not only enhances the app-building process by reducing cognitive friction and improving collaboration between design and engineering but also lays the groundwork for future advancements like instant theming and smarter debugging workflows. This initiative reflects Lovable's commitment to transforming software development and invites passionate engineers to join their innovative team.

Lovable has introduced Versioning V2.0, a significant update aimed at enhancing the speed, stability, and control for users building AI-driven applications. This update provides greater flexibility, better tracking, and an improved workflow through features like bookmarks for saving key edits, a more organized version history similar to Google Docs, and smarter restore functions that create new edit cards akin to git revert commits. The update is designed to be seamless, intuitive, and reliable, addressing user requests for a better user experience. Lovable plans to continue refining the versioning workflow to further enhance user control and creativity.

Vibe coding, introduced by AI researcher Andrej Karpathy, represents a significant shift in software development by allowing users to create applications using natural language descriptions while AI generates the actual code. This approach democratizes app development, making it accessible to non-developers, and significantly speeds up the process by reducing the need for manual coding and debugging. Lovable, a company at the forefront of this technology, demonstrated its practicality by building a complete event management app in collaboration with Supabase in a livestream. While there are concerns about code quality and maintainability, Lovable addresses these with built-in validation and real-time feedback. This method not only boosts productivity for developers by letting them focus on creative problem-solving but also empowers a broader range of people to bring their ideas to life without extensive coding knowledge. As AI continues to evolve, vibe coding is poised to further integrate business logic and domain-specific knowledge, enhancing the capability to build industry-specific applications.