CHAT GPTSTEERING the AI CODING ASSISTANT

🧭 Master Class: Steering Your AI Coding Assistant

Steering is about guiding the AI step-by-step, especially within your IDE where it can see your existing code. By providing the logic "skeleton" via comments, you turn the AI into a powerful pair programmer that handles syntax while you handle architecture.

1. Inline Steering with Comments

This is the most powerful technique for real-time coding assistance. You write the "What" in comments, and the AI fills in the "How."

A. Step-by-Step Logic Guidance

def process_orders(orders):
    # First, filter out cancelled orders (status = 'cancelled')
    active_orders = [order for order in orders if order['status'] != 'cancelled']
    
    # Then, calculate total for each order (price * quantity)
    for order in active_orders:
        order['total'] = order['price'] * order['quantity']
    
    # Now, group orders by customer_id and sum their totals
    customer_totals = {}
    for order in active_orders:
        customer_id = order['customer_id']
        # If customer doesn't exist, initialize to 0
        if customer_id not in customer_totals:
            customer_totals[customer_id] = 0
        customer_totals[customer_id] += order['total']
    
    return customer_totals

B. Data Transformation Pipeline

def clean_dataset(raw_data):
    # Step 1: Remove rows with any null values
    cleaned = [row for row in raw_data if all(val is not None for val in row.values())]
    
    # Step 2: Convert "$XX.XX" strings to float
    for row in cleaned:
        if isinstance(row['price'], str) and row['price'].startswith('$'):
            row['price'] = float(row['price'].replace('$', ''))
    
    # Step 3: Normalize email addresses to lowercase
    for row in cleaned:
        if 'email' in row:
            row['email'] = row['email'].lower().strip()
            
    return cleaned

2. File-Level Steering with Context

Use docstrings and TODOs to define the "soul" of a new file before logic is written.

Creating New Utility Files (auth_utils.py)

"""
Authentication utilities for the Flask application.
Uses JWT tokens and bcrypt for password hashing.
"""
# TODO: Import necessary libraries (jwt, bcrypt, datetime)

def hash_password(password: str) -> str:
    """Hash a password using bcrypt with salt"""
    pass

def verify_jwt_token(token: str, secret: str) -> dict:
    """Verify and decode JWT token"""
    pass

3. API & Database Development Steering

Steer toward validation and security requirements within specific routes.

Express.js Route with Validation:

// POST /api/users - Create a new user
router.post('/users', async (req, res) => {
    // 1. Validate request body (name, email, password)
    // 2. Validate email format using regex
    // 3. Check if user already exists in DB
    // 4. Hash the password using bcrypt (10 rounds)
    // 5. Generate JWT token and return 201 response
});

💡 Pro Steering Techniques

• The "Fill-in-the-Blanks" Pattern: Initialize a class and use # TODO or pass to invite the AI to implement specific methods based on your configuration.
• The "Fix This Code" Pattern: Paste buggy code and add a // BUG: comment. The AI will prioritize fixing that specific issue over a generic rewrite.

Happy Prompting! — STEERING AI Coding Assistant