Code with AI

GSID Generator Project

Task: Create a Global Unique Identifier Generator (GSID)
Purpose: Explore different strategies for AI-assisted development
Original Project: metarhia/metautil/tree/gsid-ai

Overview

This project demonstrates three different approaches to AI-assisted development by implementing a Global Unique Identifier (GSID) generator. Each approach represents a different strategy for working with AI tools, from basic prompting to structured technical specifications.

Project Structure

• 0-Simple/ - Simple Math.random() implementation
  • TASKS.md; REPORT.md; Implementation: id.js
• 1-Prompt/ - Short prompt in vibe coding style
  • TASKS.md; REPORT.md; Implementation: gsid.js
• 2-Chat-steps/ - Step-by-step chat with AI in vibe coding style
  • TASKS.md; REPORT.md; Implementation: gsid.js
• 3-Tech-specs/ - Technical specification implementation
  • TASKS.md; REPORT.md; Implementation: gsid.js
• 4-By-example/ - Browser-compatible implementation
  • TASKS.md; REPORT.md; Implementation: id.mjs
• benchmark.js - Performance comparison tool
• README.md - This file

Development Strategies

1. Short Prompt Approach

Basic vibe coding without deep understanding of the task or development process.

Characteristics:

• Minimal task description
• Limited understanding of AI-generated code quality
• Basic implementation without optimization
• Focus on quick results rather than best practices

Human Developer Time: ~5-10 minutes

• Task Definition: 2 minutes (basic prompt)
• Code Review: none
• Testing: 3 minutes (basic validation)

2. Step-by-Step Chat

Iterative vibe coding with desire for improvement but limited technical depth.

Characteristics:

• Interactive development process
• Attempts to improve implementation
• Limited understanding of code quality and optimization
• Focus on incremental improvements

Human Developer Time: ~20-30 minutes

• Initial Task: 5 minutes
• Code Review & Refactoring: 5 minutes (SRP violations, unnecessary methods)
• Architecture Decisions: 5 minutes (class structure, method removal)
• Testing & Validation: 5 minutes (comprehensive testing)

3. Technical Specification

Professional development workflow with AI as a regular team member.

Characteristics:

• Detailed technical specifications
• Clear requirements and constraints
• Focus on performance, security, and best practices

Human Developer Time: ~2-3 hours

• Requirements Analysis: 30 minutes (detailed specifications)
• Technical Design: 45 minutes (performance targets)
• Implementation Planning: 30 minutes (optimization strategies)
• Testing Strategy: 20 minutes
• Performance Analysis: 30 minutes

Human Developer Time Comparison

Approach	Time	Quality	Performance	Method	Cost
Short Prompt	~5 min	Basic	1.02M IDs/sec	Basic AI prompt	$5-10
Step-by-Step Chat	~20 min	Good	1.00M IDs/sec	Iterative AI development	$10-20
Technical Specification	~2 hours	Excellent	4.65M IDs/sec	Detailed AI specifications	$150-250
Expert Traditional	~18 min	Excellent	5M+ IDs/sec	Optimized implementation	$25-50

Key Insights:

• Technical specification requires 24x more time but produces 4.6x better performance
• Step-by-step chat provides 4x time investment for moderate quality improvement
• Short prompt is fastest but produces lowest quality results
• Time investment directly correlates with code quality and performance

AI vs Traditional Development Comparison

Performance Comparison

Based on comprehensive benchmarking, the implementations show significant differences:

#	Implementation	Performance	Memory Usage	Size	Entropy	Theoretical
0	Simple	2.01M IDs/sec	93MB	24 chars	4.987 bits/char	120 bits
1	Prompt	1.27M IDs/sec	27MB	27 chars	4.776 bits/char	162 bits
2	Chat-steps	1.22M IDs/sec	39MB	27 chars	5.196 bits/char	162 bits
3	Tech-specs	5.95M IDs/sec	86MB	24 chars	6.000 bits/char	144 bits
4	ID (By-example)	1.54M IDs/sec	84MB freed	24 chars	6.000 bits/char	144 bits
-	UUID v4	3.16M IDs/sec	470MB	36 chars	4.050 bits/char	122 bits

Key Findings:

• Tech-specs approach produces the most efficient implementation - 2x faster than UUID v4
• Simple implementation shows excellent performance (2.01M IDs/sec) with minimal code complexity
• ID (By-example) provides excellent browser compatibility with good performance (1.54M IDs/sec) and compact size (24 chars)
• All implementations maintain perfect collision resistance (0% collision rate)
• Node.js 24 provides native Web Crypto API support without experimental flags

Development Tools Used

• Cursor: claude-4-sonnet, claude-3.5-sonnet, gpt-4.1, o3, gemini-2.5-pro
• VS Code with Copilot: gpt-4.1, claude-4-sonnet, gemini-2.5
• ChatGPT: gpt-4o, o4-mini-high, gpt-4.1, gpt-4.5

Getting Started

1. Clone the repository: git clone <repository-url>
2. Open folder: cd CodeWithAI
3. Install dependencies: npm i
4. Run tests: npm t
5. Run benchmark: node benchmark.js

Implementation Details

4-By-example (Browser-Compatible)

The browser-compatible implementation demonstrates adapting Node.js code for web environments:

• Technology: ES Modules with Web Crypto API
• Key Features:
  • Pure Web API usage (crypto.getRandomValues())
  • No Node.js dependencies
  • Cross-platform compatibility
  • Same character set and algorithm as Tech-specs
• Performance: 1.54M IDs/second with 0% collision rate
• Size: 24 characters (same as Tech-specs)
• Usage:

  // Browser
  import { generateID } from './id.mjs';
  const id = generateID();
  
  // Node.js (native support in v24+)
  node benchmark.js

Lessons Learned

• Technical specifications produce the most efficient and maintainable code with AI
• Iterative development can improve results a little but requires technical expertise
• Basic prompting often leads to functional but suboptimal implementations

License

This project is part of the original metarhia/metautil repository and follows its licensing terms.