Idea /

GEN-3D-ASSETS Project

GEN-3D-ASSETS as a vibe project in an unfamiliar domain to test the limits of AI coding and my own ability

GEN-3D-ASSETS was the project I used to test the limits of AI plus my own skills in an unfamiliar domain. It was created on August 26, 2025. I no longer subscribe to the Meshy API, so the project has been shut down.

Project Repository

https://github.com/Lemonadeccc/GEN-3D-ASSESTS

Page Preview

image.png

image1.png

I. Project Boundaries

This was an NFT project used as a first pass at Web3. The intended product flow was "generate -> preview -> select", forming a complete NFT platform.

The technical direction was Next.js + R3F + Solidity + Vercel + Claude.

II. Technical Constraints

In practical use, loading obj / fbx / gltf should not depend on a Next.js proxy, and models or textures should not be loaded through blob.

  1. Model assets should be distributed in a way browsers can consume directly
  2. Full state feedback and comfortable interaction support are required
  3. Multi-3D-model performance issues need to be handled

III. Engineering

The engineering priorities can be compressed into a few points:

  1. Decide the contract and chain stack first, such as Solidity / Foundry / Hardhat
  2. Tighten commit and code conventions first with commitlint, eslint, and husky
  3. Use GitHub Actions for CI/CD, combine Qodo for PR review, and use PocketFlow to generate repository docs
  4. Leave room for performance optimization, focusing on model loading and Turbopack incremental updates
  5. Prioritize the MVP first, then iterate on GSAP animation and static asset optimization

IV. Business Flow

  1. Use meshy.ai or another generative API to create 3D assets
  2. Decide whether generation results come back through polling or some other mechanism
  3. Design the interaction around preview, selection, and waiting after the model is returned
  4. After the user picks a result, enter the on-chain contract flow
  5. The deployment approach has to be considered together with the engineering constraints
  6. Pay attention to API key storage and security

V. Project Value

The platform goal was to generate NFT-like personalized 3D assets: based on a user's profile and prompt preferences, output unique non-fungible 3D results. Or upload 3D assets on-chain for intellectual property management. My personal goal was to explore the limits of AI coding and my own ability in an unfamiliar domain. The practical conclusion was that AI coding is not very workable in a completely unfamiliar domain. With the AI capability available at that time, it could not guide me from my own knowledge base all the way to a real production-ready launch.