Nahj AI Assistant

How can I help you today?

Ask me anything about software development

Quick start ideas
HomeFrontendBackendArch & InfraDeployAI

AI-Assisted Coding

Leverage AI tools effectively

AI coding assistants can dramatically speed up development when used correctly. This section covers environment setup, which AI models to use for different tasks, subscription recommendations based on your budget, and best practices to get the most out of AI tools.

Windows Setup

If you are coding on Windows, install WSL (Windows Subsystem for Linux) and put your projects there.
Keep projects in Linux filesystem (~/), not Windows (/mnt/c/). Windows paths are 10x slower.

AI Models

No single model excels at everything. Use multiple models for different tasks.
Frontend
Gemini Pro 3 excels at frontend code but struggles with backend.
Backend
Claude Opus 4.5 is best. Sonnet 4.5 and GLM 4.7 are good alternatives.

What is Your Budget?

Monthly Budget

$40/month

Free
$10
$20
$40
$100
$200

Free

Free Tools

• ANTIGRAVITY IDE: Gemini Pro 3 & Flash 3.0 • Qwen Code: backend tasks • Great for learning and small projects

$10/month

Pick One

Trae

• Trae IDE: Claude Sonnet 4 ($10) • Z.ai Lite ($6): Use GLM 4.7 in Claude Code • GitHub Copilot (free for students)

$20/month

Codex CLI

• OpenAI Codex CLI • High usage limits • Good for GPT models

$40/month

Claude Code + Codex CLI

• Claude Code + Codex CLI ($20 each) • Best combo for most developers

$100/month

Claude Code Max 5x

• 5x higher rate limits • Best for most projects, hard to reach limits

$200/month

Claude Code Max 20x

• 20x highest rate limits • Maximum power for intensive projects • Only if $100 plan does not fit your limits

Why to Avoid AI IDEs like Cursor?
Avoid
AI IDEs like Cursor give you roughly $20 worth of API usage per month. You could burn through it all in a single day, leaving you with nothing for the rest of the month. Services like Codex or Claude Pro reset limits weekly, so you always have something to work with.

Best Practices

Simple Tasks
Use lightweight models (Haiku, GPT Mini) for color changes and minor edits to save your usage limits.
Complex Features
NEVER implement directly. ALWAYS plan first. Build frontend completely before backend.
Memory Files
Maintain context files for AI agents to understand your project.
Provide Context
Ask about best practices and providers before implementing new features.
Iterate
You will not get perfect results in one prompt. Expect multiple revisions.
Security Review
Use /security-review before pushing to main branch (Claude Code slash command).
Avoid MCP
Avoid
Do not use MCP, it adds excessive context to the AI, increasing token usage and reducing response quality.
Do Not Vibe Code
Avoid
Do not be a vibe coder, read and understand the code.

Learning Channels

These YouTube channels provide hands-on tutorials for AI-assisted development.

Designing Websites with AI
Learn how to design and build websites using AI tools. A hands-on tutorial showing AI-assisted web design workflows. Check out the channel for more AI development content.
Watch on YouTube
Riley Brown AI
Channel
Builds complete AI applications from scratch in each video, walking through the entire development process. Learn practical AI app development by coding along with real projects.
Visit Channel
AI Code King
Channel
Covers AI tools, use cases, and model comparisons with benchmarks to help you choose the best models for your needs. Stay updated on the latest AI coding developments.
Visit Channel
Robin Ebers
Channel
Creates tutorials for multiple AI tools with detailed explanations while building projects. Excellent resource for understanding different AI platforms and their practical applications.
Visit Channel