AI Content Assistant
An AI content tool for small business owners. Upload a photo and write a short note — the AI generates a blog post, SNS copy, short-form script, and hashtags. The visual prototype is complete. The next stage is building a fully functional MVP with real AI integration.
Development record of an AI content tool MVP for small business owners
What you'll find here
- Whether a local shop owner with no social media skills can make content with an AI tool — the validation process
- How an idea became a live service in one day: from wireframe to Cloudflare deployment
- Where MVP scope was cut, and what was decided to validate first
- How the Claude Design → Claude Code Handoff workflow performed in a real build
Current Status
MVP live service deployed (2026-04-24). View AI Content Assistant Prototype. Changes go live 23 seconds after git push.
The visual prototype is live. AI feature integration and full MVP development are in progress.
Next: send the link to 5 real small business owners and collect usage feedback. A tool that works is not the same as a tool that gets used.
Who This Tool Is For
Local shop owners. Between cooking, serving customers, and managing inventory, there's no time left to post on social media. Outsourcing content is too expensive. There's no one to ask for help writing copy.
The tool's single assumption: one photo and a short note is enough. It should be usable without any writing ability or social media intuition.
The Problem Being Solved
The content creation cost for small business marketing — not money, but time and skill.
They want to post about a new menu item but don't know how to write the caption. They have a photo but don't know what hashtags to use. They need different copy for Instagram and their blog but don't have time to write both. These three barriers are what this tool is built to remove.
Current MVP Scope and How It Works
8 screens. Upload a photo and a short memo — the OpenAI GPT-4o API generates content using structured prompts tailored to the business type and content goal.
| Input | Output |
|---|---|
| Photo + short memo | Blog post (long-form) |
| Business type (café · hair salon · craft studio · retail) | SNS copy (Instagram · Facebook) |
| Content goal (new product · grand opening · event · return visit) | Short-form script (Reels · Shorts) |
| Hashtags (by category and location) |
Infrastructure: Cloudflare Pages (frontend) + Worker (API proxy), GitHub Actions auto-deploy, current operating cost ₩0.
What's Been Built
- Prototype v1 → v2 (2026-04-22) — Mobile web prototype built with Claude Design. Input-to-output flow redesigned based on v1 feedback
- MVP live service deployed (2026-04-24) — Next.js + TypeScript + Tailwind CSS, Cloudflare Pages/Worker pipeline, 8 screens
- Deployment automation — GitHub Actions deploys Pages and Worker simultaneously; git push → live in 23 seconds
- Current infrastructure cost: ₩0
Recent Updates
- 2026-04-24 Next.js MVP deployed to Cloudflare. OpenAI GPT-4o API connected via Worker, GitHub Actions auto-deploy pipeline built
- 2026-04-22 · v2 Claude Design prototype v2. Input-to-output flow redesigned, UI brought to service-level polish, MVP scope finalized
Featured Logs
- AI Content Assistant — From Idea to Screen to Implementation Scope An idea written in a business plan application became an interactive mobile prototype in one day. The MVP scope only became visible once there was a screen to look at.
- AI Content Assistant MVP — From Dev Setup to Live Service in One Day From VS Code install to Cloudflare deployment in one day. 23 seconds from git push to production.
Featured Insights
- Ideas Are Validated Faster with Screens Than with Words Building a screen is faster and more accurate than writing a spec. The screen reveals what the spec missed.
- No Coding Skills, Live in One Day — Why Claude Design + Claude Code Made It Possible AI closed the gap between design and development. That's different from "AI writes the code for you."
Next Validation Steps
- Send
aicontent.chulbuji.comlink to 5 real small business owners → collect usage feedback - First usability check: "Did you immediately understand what to do on this screen?"
- Review generated content quality by business type and goal against real-user standards
- Simplify UI copy — rewrite from the small business owner's perspective