Vibe Coding Conversation Recap

Introduction

I wanted to explore vibe‑coding a web app with Loveable to track my projects. The conversation with Claude helped flesh out the core ideas, the Kanban workflow, and a systematic building process.

Core Ideas for the App Tracker

• App metadata: name, description, category (web, iOS, Android, cross‑platform).
• Build status: idea, in progress, launched, paused, abandoned.
• AI tool used: Loveable, Bolt, Cursor, v0, etc.
• Tech stack / framework notes.
• Links: live URL, GitHub repo, app store listing.
• Progress & activity: session log, changelog, time spent, last active date.
• Metrics: user count, revenue, feedback, usage toggle.
• Ideas pipeline: backlog of features, priority ranking, blockers.
• Discovery & reflection: tags, learning notes, public/private toggle.

Kanban Overview

A visual board with columns: - Idea → Building → Testing → Launched → Paused → Abandoned Each app appears as a card showing name, platform, AI tool, last update, and a status tag.

Systematic Building Process

1. Define the idea – problem, target user, core feature, scope.
2. Write the prompt – detailed prompt for Loveable.
3. Generate the first version.
4. Review & list fixes – intentional, not random.
5. Iterate in focused rounds – layout → features → polish.
6. Test – especially on mobile.
7. Deploy / share.
8. Collect feedback – decide next steps.

Guided AI Assistant Concept

Each stage gets its own chat bot: - Idea stage – asks about problem, user, core feature, one‑sentence description. - Build stage – helps craft the Loveable prompt. - Testing stage – provides a checklist and helps write fix prompts. - Launch stage – assists with description and sharing.

Cost Considerations

• Use Claude Haiku for cheap chat.
• Keep conversations focused, limit messages per stage.
• Summarise after each stage.
• Hosting the chatbot on your own Hermes server makes it essentially free.

Next Steps & Ideas

• Implement the guided assistant as a web component within the tracker.
• Store each stage’s transcript in the FastMCP SQLite DB.
• Add a “cost estimate” badge showing projected chat spend.
• Enable sharing of completed idea definitions via a PDF export.
• Explore integrating a lightweight LLM (e.g., Ollama) for offline use.

---

Generated from the Claude conversation on May 2026.