DEV Community

I. Yosun Chang
I. Yosun Chang

Posted on

Project Cameo-Ecosystem aka Merchmatica

#showdev #ai #startup #aws

Inspiration

This was a Deep End of the Pool experiment: could we vibe-code a complex, enterprise-grade, multi-value-prop AI ecommerce platform with no manual code fixes—only plugging AWS credentials?

Merchmatica is also inspired by Stephen Wolfram’s *A New Kind of Science.

Where Wolfram showed that *simple local rules in cellular automata can generate infinite complexity, here we ask:

What happens when simple licensing + payment rules generate an entire merch economy?

  • A creator sets a royalty %
  • A fan generates an AI design
  • A store curates products
  • A Stripe split propagates payouts

From these atomic rules emerges a new kind of commerce: unpredictable, generative, and computationally irreducible.

What it does

Merchmatica (aka Project cameo-ecosystem — named after my dog and AI design collaborator @corgi.cam) is a platform for an emergent merch ecosystem.

  • Creators directly license and monetize their likeness for AI-downstream content creation that flows into physical merch.
  • Fans can spin up their own stores, generating apparel, postcards, even 3D-printed figurines from licensed creator IP.
  • Payments & fulfillment are handled through Stripe Connect and merch APIs, turning fan-inspired generated creations into tangible goods.

Instead of a top-down catalog, Merchmatica works like a commerce automaton: creators, fans, and stores play out simple rules that give rise to complex global markets.

How we built it

We started with a single prompt:

“Create Spec: Cameo Ecosystem: allow fans to generate brand-true images and products such as postcards, apparel and 3D printed figurines (2D/3D) from Creators, then sell products via stores.”

Then let the Kiro agent autopilot it through:

  • Spec generation + architecture by Kiro
  • Implementation assisted by VSCode Copilot
  • Stack: Next.js SSR + AWS Amplify + Stripe Connect + FAL/Replicate for AI gen
  • Multi-layered environment loader for secrets + debug endpoints for production validation

Like cellular automata rules encoded in Mathematica, Kiro’s “spec → system” process became its own rule-based generator of architecture and code.

Challenges we ran into

Deploying a Next.js SSR app on AWS Amplify exposed edge cases reminiscent of computational irreducibility—small rule-changes, big unpredictable outcomes:

  • process.env proved unreliable for SSR/API routes → secrets vanished at runtime.
  • Workarounds: multi-source loader pulling from process.env, static refs, AWS SSM, encrypted build artifacts, and .next/env.json.
  • All sensitive ops forced into Node-only space to avoid leaking secrets.
  • Added /api/debug/env + /api/debug/creators-direct to confirm runtime connectivity.
  • Wrestled with URL-encoded secrets (%24 vs $) breaking database URLs.

Takeaway: Amplify + SSR = a nontrivial automaton; resilience requires layered env logic and runtime validation.

Accomplishments we’re proud of

  • End-to-end emergent flow: licensing → AI generation → storefront → payment split → fulfillment.
  • Royalty-aware Stripe Connect flow that mirrors conservation laws in physics: nothing lost, everything distributed.
  • Debug + resilience infrastructure that exposed hidden rules of the system before users hit them.
  • Building business-grade reliability entirely through rule-based AI orchestration.

What we learned

Conversation strategies with Kiro that worked like defining rules in an automaton:

  1. Context-Rich Requests

    • ❌ Bad: “Create a payment system”
    • ✅ Good: “Build a Stripe Connect system where creators set royalty rates (0–50%), store owners curate products, fans purchase merchandise, and all parties get paid automatically with full audit trails.”

  2. Constraint-Driven Development

    • Must handle concurrency, real-time status, multi-currency, regulation compliance, and API failure recovery.

  3. Integration-First Thinking

    • Payments tied into licensing, catalog, webhook infra, and AWS monitoring from day one.

  4. Iterative Refinement

    • Started with working checkout → layered in watermark removal, email confirmations, retries, analytics.

💡 Why this worked:

  • Kiro thinks in systems, not snippets.
  • Operates with a production mindset: error handling, tests, edge cases.
  • Tracks integration surfaces across the stack.
  • Builds iteratively, carrying forward context like evolving states in a cellular automaton.

Like Wolfram’s NKS, the lesson is: simple rule-sets are enough to create worlds of complexity. Merchmatica is one such world—commerce as computation.

What’s next for Merchmatica

🚀 LAUNCH!

Top comments (0)