Solving everyday problems with vibe coding #2

🚩 Problem

There are so many AI tools and features available today that it’s often unclear which one to use. This leads to wasted time, confusion, and decision fatigue.

💡 Solution

Help users choose the right AI tool by asking a few quick questions and suggesting the most relevant options.

⛏️ Tools used

  • PRD: ChatGPT (Deep research)
  • AI Agent: Replit
  • API: Google Gemini API

Try FindMyAI 👉 https://find-my-ai-2nd-brain.replit.app

📝 Notes from the build

1. Writing the PRD with ChatGPT

  • Writing the PRD with ChatGPT felt smoother than with Google AI Studio.
  • Google AI Studio mostly paused after each step to ask simple questions like “Do you have changes?” instead of actively helping refine the PRD.
  • ChatGPT, on the other hand, highlighted important points proactively and asked for clarification. This made the collaboration feel more natural.
  • Before moving to the next step, ChatGPT summarized what it had learned from the previous discussion and explained how it would proceed. This structure was very helpful.

2. Building and iterating with Replit

  • Replit provided a much better UI editing experience compared to Lovable. It felt more like working directly in a design tool such as Figma, with intuitive control over layout and components.
  • However, although FindMyAI worked smoothly in the Replit development environment, it consistently failed after deployment. One persistent issue was fetching and displaying logo images for each tool.
  • Overall, Replit is more comfortable for people who know how to code. It is not yet suitable for non-coders who need a reliable deployment-ready solution. Without some computer science fundamentals, Replit does not offer a full end-to-end experience.

3. API integration

  • Connecting with the Google Gemini API was quick and straightforward.
  • However, I noticed that text length and tone do vary between APIs. If concise and polished text output is the goal, APIs like OpenAI or Claude may be more effective.

🌱  Takeaways

What worked

  • Collaborating with ChatGPT for the PRD clarified the problem and kept progress structured.
  • Replit’s UI editing tools made interface design more intuitive and felt closer to design-first tools.

Challenges I faced

  • Deployment on Replit was little bit unstable.
  • Text prompts alone were not enough for precise UI control.

What could be improved

  • Replit needs stronger deployment support to be viable for non-coders.
  • Different APIs should be benchmarked not only for accuracy but also for clarity and tone of responses.