Skip to content

Technical Moat

Why this is hard to replicate

1. Edge-native architecture

graph LR
    U[User] -->|"<50ms"| E[Cloudflare Edge]
    E --> W[Workers]
    E --> D1[D1 Database]
    E --> R2[R2 Storage]
    E --> DO[Durable Objects]
    E --> KV[KV Cache]

Everything runs at the edge. No origin servers. No cold starts. Sub-50ms responses globally. This isn't bolted-on CDN caching — the application logic itself executes in 300+ data centers.

2. Compliance-as-code

Every item in every store passes automated quality gates before publishing:

  • Up to 42 checks (brand, security, performance, accessibility)
  • Automated viewport testing (12 device sizes via Playwright)
  • Bundle size budgets (300KB gzipped max)
  • No-tracking enforcement (blocks Google Analytics, Mixpanel, etc.)
  • Brand consistency (fonts, colors, spacing enforced by CI)

Competitors can copy the store concept but not the quality infrastructure built into every deploy pipeline.

3. AI-native publishing

Three creation paths, all AI-powered:

Path How Time to publish
VibeCode (AI builder) Describe what you want in chat ~2 minutes
CLI Scaffold template, code, fas publish ~30 minutes
MCP (AI editor) Build in Cursor/Claude, tools deploy for you ~10 minutes

The AI doesn't just assist — it's a primary operator. FWS builds complete websites from a conversation. FAGS creates browser AI tools. PAGS creates server agents.

4. Zero-cost free tier (permanent)

Free stores cost almost nothing to operate: - Cloudflare Workers free tier: 100k requests/day - R2: first 10GB free, $0.015/GB after - D1: first 5M rows free - GitHub Actions: free for public repos

This means free stores can stay free indefinitely without VC-funded burn rate. They're self-sustaining.

5. Shared infrastructure, independent products

graph TD
    subgraph "Shared (built once)"
        Auth[GitHub/Google OAuth]
        Host[R2 Host Workers]
        CI[CI/CD Pipelines]
        DS[Design System]
        Secrets[Doppler Secrets]
    end

    subgraph "Independent (per-store)"
        SDK1[App SDK]
        SDK2[Game SDK]
        SDK3[Agent SDK]
        SDK4[CMS Engine]
    end

    Auth --> SDK1
    Auth --> SDK2
    Auth --> SDK3
    Host --> SDK1
    Host --> SDK2
    CI --> SDK1
    CI --> SDK2
    CI --> SDK3
    CI --> SDK4
    DS --> SDK1
    DS --> SDK2
    DS --> SDK3

Build once (auth, hosting, CI, secrets, design system), deploy across all 8 stores. Each store's SDK is independent but the infrastructure cost is amortized.

6. Ecosystem lock-in (positive)

  • Creators use the same identity across all stores
  • One CLI (fas) works for apps; fgs for games; same mental model
  • Same design system = users feel at home across stores
  • Same compliance = quality is uniform
  • Cross-store MCP = AI agents can operate across the ecosystem

7. AI model portability (FAGS unique moat)

Browser-based AI tools run on the user's hardware. This means: - Zero inference cost to the platform - Works offline (after model download) - No API rate limits - Privacy (data never leaves the device) - Scales infinitely (more users = more GPUs, not more servers)

No server-side AI company can compete on cost with "runs on user's GPU for free."