Thumbnail
HomePage
What you can do
Analysis
Visualize in 3D
Chat
Code

🚀 About CodexOrb

CodexOrb is an AI-powered collaborative coding platform designed to bridge the gap between natural language and real-time software development. The project was created for the Bolt.new One Prompt Challenge with a bold mission: turn conversations into functional, optimized code — visually and collaboratively.

🌟 Inspiration

We were deeply inspired by the increasing friction developers face when switching between tools like ChatGPT, IDEs, GitHub, and deployment platforms. The magic of tools like Raycast, GitHub Copilot, Vercel, and Linear showed us that developer experience can be beautiful, fast, and intelligent.

Yet, we felt something was missing:

A space where developers could talk to AI, visualize code as an evolving 3D system, and collaboratively build full-stack apps — in one flow.

Thus, CodexOrb was born:
An intelligent, visual, team-friendly code companion built to feel like magic.

🧠 What We Learned

Throughout the build, we learned:

How to leverage Bolt.new’s AI agent system to build production-grade applications from a single smart prompt.
Techniques to generate structured, collaborative interfaces through natural language.
The challenge of aligning AI-generated code with real-world development best practices, especially in multi-turn prompts.
How to balance developer control with AI automation — letting users edit, update, debug, and deploy, all from a conversation.

🏗️ How We Built It

We crafted a single, highly detailed Bolt.new prompt that generated:

A chat interface with dual-mode toggles (General Chat & Code Instruction)
AI logic to:
- Generate code from prompts
- Update existing files
- Fix bugs and explain changes
A 3D code sculpture using WebGL or Three.js where:
- Nodes = functions/files
- Edges = dependencies
- Red = bugs, Green = optimizations
A clean UI inspired by Vercel, Linear, and GitHub:
- Dark/light toggle
- Inter font and monospaced elements
- Minimal 3D canvas + chat side panel
Deployment powered via Supabase (backend) and Vercel (frontend)

All generated automatically by Bolt.new using a deeply engineered single prompt.

🚧 Challenges We Faced

Context retention: Making sure AI-generated code recognized prior structure and edits
Visualization accuracy: Mapping code into 3D space meaningfully
Prompt compression: Fitting all product logic, UI detail, backend scaffolding, and interactivity into one coherent Bolt.new prompt
Instruction clarity: Teaching the AI how to separate chat from coding actions using a toggle system
Time limits: Building something this ambitious within the hackathon window