[Lovable] Improve stability and reduce bugs in high-fidelity prototypes

At work, I use Lovable not to build production-ready apps or hand off to developers (It will only give them a headache). Instead, I use it to create realistic prototypes that we can test with real users within the organization. It helps me validate ideas quickly without getting bogged down in details in Figma too early. Here’s how I iterate: 🖥️ Rough Design in Figma. A lot of the work I do involves existing designs of enterprise tools, so I usually start with a rough draft in Figma just to guide my thinking. 📜 Drafting Prompts in Notion. I start by outlining the key interactions and describing the UI I want to build. I usually jot these down in Notion, since ideas often come to me while commuting or when something comes to mind during work. 🌟 Refining Prompts in ChatGPT. I use ChatGPT to refine my prompts by either improving their structure (typically in markdown format) or by prompting it to ask clarifying questions about key interactions, design guidelines, and edge cases I might have missed. 🏗️ Base Prototype in Lovable. I spend some time in Lovable to build the base prototype, which includes the main features and UI. Depending on the complexity, it usually costs me 100 to 150 credits, including the five daily free credits (yes, I make the most of those!). 🔀 Remix Variants for Testing. Once I’m happy with the base prototype, I remix it into different variants depending on the ideas we want to test. Including history when remixing is important to keep context, and I’ve found it helpful for referring back to previous iterations. Using the remixed prototypes, I run moderated tests with internal users and gather insights from their interactions and feedback. Watching users engage with something that feels real makes it easier to align with stakeholders and decide what to take forward. Since I am working on existing enterprise apps at Agoda, I don’t treat Lovable’s code as something we’ll ever ship. It’s just a way to figure things out before committing in Figma. When I return to Figma, there’s less guesswork and less moving pixels around. It feels more like assembling components like LEGO with clear instructions already in front of me. I just copy the validated states and flows, then prepare everything for handoff.