· technical  · 3 min read

What I Learned Building 99 Minds: Idea Capture for People with Too Many Ideas

DRAFT

Outline

Hook: I built 99 Minds because I was drowning in my own ideas. Voice notes scattered across apps, notebooks full of half-thoughts, brilliant shower epiphanies forgotten by the time I found my phone. The problem wasn’t lack of ideas—it was that my ideas were dying because my capture system sucked.

Core Argument: Most productivity tools fail creative people because they’re designed for task completion, not idea preservation. 99 Minds taught me that the best idea capture system needs three things: friction-free input, intelligent organization, and the ability to resurface ideas at the right moment.

Key Sections:

  1. The Problem: Creative Minds Are Idea Generators, Not Idea Organizers

    • The “creative curse”: 10 ideas before breakfast, 0 executed by dinner
    • Why traditional productivity tools fail: too structured, too slow
    • The “idea graveyard” phenomenon
    • User research: talked to 50+ creative professionals about their systems

  2. Design Principle #1: Capture Must Be Faster Than Thought

    • Voice-first interface: speak while walking, driving, showering
    • No forms, no categories upfront, no friction
    • Technical challenge: real-time transcription that actually works
    • Why I chose Whisper API + custom post-processing
    • Mobile-first: most ideas happen away from desk

  3. Design Principle #2: AI Should Organize, Not Decide

    • Auto-categorization using embeddings + clustering
    • Tagging suggestions, not forced taxonomies
    • Link detection: “This sounds like that other idea from last week”
    • The danger of over-automation: users need to feel ownership
    • Balance: helpful without being presumptuous

  4. Design Principle #3: Resurfacing Is Harder Than Storing

    • Weekly “idea reviews” with AI-generated summaries
    • Context-aware suggestions: “You’re working on X, remember Y?”
    • Similarity search: find related ideas across time
    • The “serendipity engine”: random idea prompts
    • Metrics that matter: ideas executed, not just captured

  5. What Went Wrong (And What I Fixed)

    • V1: Too much AI, felt impersonal → scaled back automation
    • Search was too literal → added semantic search
    • Users wanted privacy → local-first options, encryption
    • Monetization struggle → freemium model with power features
    • The feature bloat trap: said no to 100 feature requests

  6. The Bigger Picture: Building for Yourself First

    • Dogfooding: I’m the primary user
    • Why “scratch your own itch” works for niche products
    • Finding your 1,000 true users
    • When to pivot vs. when to persist

Examples/Stories:

  • Origin story: Missing a great app idea because I couldn’t find my note
  • User story: A writer who captured entire novel plot while commuting
  • Technical failure: Early transcription quality issues, angry users
  • Business lesson: First 100 users came from Twitter thread, not ads
  • Personal impact: How 99 Minds helped me ship more projects

Takeaways:

  • Build for the moment of capture, not the moment of review
  • AI should feel like a helpful assistant, not an overbearing manager
  • Voice interfaces remove friction but require technical investment
  • The best product insights come from using your own product daily
  • Niche products can thrive—you don’t need millions of users

Cross-Links:

  • ← “From Mixtapes to Machine Learning” (Series 1-1)
  • → “RAG, But Make It Real Life” (Series 1-4)
  • → “The 99 Minds Principle” (Series 3-26)
  • → “How I Decide If an Idea Is an App” (Series 2-13)
  • → “The Narrow But Complete Rule” (Series 2-12)
Share:
Back to Blog