Planning the Project with AI

While “vibe coding” (building an app with a single click) is trendy, I wanted to approach this project more methodically. So I gathered various ideas from my head and asked Claude to create a “100-day development challenge” plan.

By the way, I prefer to use English when conversing with LLMs. Although recent models handle Korean queries naturally, considering the ratio of training data by language, I believe English performance is still superior.

Anyway, Claude developed a 100-day “building in public” challenge plan considering these factors:

  • Idea conception and POC, domain and brand naming
  • Building the website and generating content using AI
  • Multi-language support (English/Korean), visitor analytics, advertising, social marketing
  • A side project with maximum 2-3 hours of work per day

After several rounds of my additional feedback and Claude’s revisions, we refined the plan to version 29(!) Here’s the condensed version of the 100-day plan:

  • Phase 1: Foundation and MVP Development (Days 1-30)
    • Weeks 1-2: Detailed planning, brand and domain naming, initial static site implementation
    • Weeks 3-4: Core feature development, story content generation workflow, MVP launch with one demo content
  • Phase 2: Story Content Automation and Viral (Social Sharing) Features (Days 31-60)
    • Weeks 5-6: Generate second story pack with automated workflow, implement social sharing features
    • Weeks 7-8: Content expansion and MBTI-based personality test concept implementation
  • Phase 3: Feature Enhancement and Scaling (Days 61-80)
    • Weeks 9-10: User account implementation, background music addition
    • Weeks 11-12: Commercialization and user analytics
  • Phase 4: Community Building and Growth (Days 81-100)
    • Weeks 13-15: Social content marketing focused on short-form videos

Leveraging MBTI Personality Types for Viral Marketing

“Interactive stories based on classic novels” alone seemed insufficient to spark people’s interest. Then I thought, “What if we incorporate MBTI?”

As shown in this example generated with Gemini Nano (banana model), we score MBTI traits based on user choices at each story branch and display their personality type on the ending screen.

Of course, we can’t expect the accuracy of an actual test—this is purely for engagement and viral potential. I confirmed that the term “MBTI” and official test materials can’t be used without permission from the copyright holder.

However, the 16 personality types and their abbreviations (INTP, ENFP, etc.) are not copyrighted and can be freely used by anyone.

In conclusion, as long as we don’t explicitly mention the term “MBTI,” my ending screen concept has no trademark or copyright issues.

Refining Ideas Through AI Conversation

Beyond what I’ve written above, through conversations with Claude, I refined thoughts and plans on these topics:

  • Mobile app development: For the challenge phase, support only mobile web instead of native apps
  • Content length strategy: Start with “mini-story” versions under 3 minutes, then expand to “full-story” versions later
  • Content factory approach: Provide pre-generated “content packs” statically instead of real-time content generation

There was a recent issue where ChatGPT’s model developed an anti-pattern of being too sycophantic to users. I was also feeling tempted to keep having these conversations because Claude is so accommodating.

But if I keep just refining plans, I’ll fall into “analysis paralysis”—planning everything but starting nothing. It’s time to move to the next stage.

To be continued…