WEB DEV

Appearance

Prompt Engineering: Talking to the Junior Dev 🤖🗣️ ​

Imagine you have a very fast, very eager, but slightly literal-minded Junior Developer sitting next to you.

  • If you say "Fix this code," they might delete everything.
  • If you say "Make it better," they might add emojis everywhere.
  • If you give clear, specific instructions, they will do exactly what you want in seconds.

Prompt Engineering is just the skill of giving good instructions to this Junior Dev.

The 3 Golden Rules 🌟 ​

1. Be Specific (Context is King) 👑 ​

The AI doesn't know your project structure, your variable names, or your life story unless you tell it.

Bad Prompt ❌Good Prompt ✅
"Write a function to read a file.""Write a Python function using pathlib that reads a JSON file and returns a dictionary. Handle the FileNotFoundError."

2. Give Examples (Few-Shot Learning) 🎯 ​

Show, don't just tell. If you want a specific output format, give an example.

Prompt:

"Convert these names to email addresses. Example: 'John Doe' -> 'john.doe@company.com' Input: 'Alice Smith'"

3. Iterate (The Conversation) 🔄 ​

You rarely get the perfect code on the first try. Treat it like a chat.

  1. Ask: "Write a script to scrape a website."
  2. Refine: "Okay, but use BeautifulSoup instead of Selenium."
  3. Fix: "It's crashing on line 10. Here is the error..."

Advanced Techniques for Beginners 🚀 ​

The "Persona" Pattern 🎭 ​

Tell the AI who it should be. This sets the tone and quality of the response.

Prompt:

"Act as a Senior Python Backend Engineer. Explain Python Decorators to me like I am a 10-year-old."

Chain of Thought (Think Step-by-Step) 🧠 ​

If you ask a complex logic question, the AI might guess. Force it to show its work.

Prompt:

"I have a list of numbers: [10, 2, 5, -1, 0]. I want to find the second largest number. Think step-by-step and explain your logic before writing the code."

Use Delimiters 🚧 ​

Use triple backticks (```) to separate your instructions from your code. This helps the AI understand what is what.

Prompt:

Refactor the code below to use List Comprehensions.

python
numbers = [1, 2, 3]
squares = []
for n in numbers:
    squares.append(n * n)

The "ELI5" Pattern (The Teacher) 👶 ​

If a concept is confusing, ask the AI to simplify it.

Prompt:

"Explain Python recursion to me like I am 5 years old. Use an analogy involving Russian Nesting Dolls (Matryoshka)."

The "Style Check" Pattern (The Linter) 🧹 ​

You can use AI to check your code against style guides without installing tools.

Prompt:

"Review this code for PEP-8 compliance. Point out any bad variable names or formatting issues."

The "Data Formatter" Pattern (The Intern) 📋 ​

You have messy text and want clean data structures.

Prompt:

"I have this raw string of names and ages: 'John:25, Alice:30, Bob:22'. Convert this into a Python list of dictionaries like [{'name': 'John', 'age': 25}, ...]."

The "Regex Wizard" Pattern 🧙‍♂️ ​

Regular Expressions are hard. Let AI write them.

Prompt:

"Write a Python function using re module to extract all phone numbers from a text. Format matches: (123) 456-7890."

The "Syntax Fixer" Pattern (The Typos) 🛠️ ​

Python hates messy indentation. AI can spot it instantly.

Prompt:

"I am getting an IndentationError. Fix this code for me.

python
def hello():
print("Hi")
```"

The "Logic Translator" Pattern (English to Code) 🗣️ ​

Struggling to write the if statements? Describe it in plain English.

Prompt:

"Write a Python function check_access(age, has_ticket) based on this logic: If they are under 18, reject them. If they are 18 or older but don't have a ticket, reject them. Only let them in if they are 18+ and have a ticket."

Result:

python
def check_access(age, has_ticket):
    if age >= 18 and has_ticket:
        return True
    return False

The "Skeleton Builder" Pattern (Scaffolding) 🏗️ ​

Don't know where to start? Ask for the structure, not the full code.

Prompt:

"Write the skeleton for a Python script that reads a CSV file and calculates the average. Define the functions I need (like read_file, calculate_avg) but use pass for the body."

Result:

python
def read_file(filepath):
    # TODO: Open file and return data
    pass

def calculate_avg(data):
    # TODO: Calculate average
    pass

def main():
    data = read_file("data.csv")
    print(calculate_avg(data))

The Anatomy of a Perfect Prompt 🧬 ​

A perfect prompt has 4 parts:

  1. Role: Who is the AI? ("Act as a Senior Python Engineer...")
  2. Task: What should it do? ("Write a unit test...")
  3. Context: What constraints? ("Using pytest, focusing on edge cases...")
  4. Output: How should it look? ("Output only the code block, no explanation.")

Task: The "Bad" Prompt

Try asking ChatGPT: "Write code." See what happens. It will likely ask "What kind of code?" or write a generic "Hello World".

Example: The "Iterative" Game 🎮 ​

Let's say you want to build a Number Guessing Game.

Round 1 (The Draft):

"Write a Python number guessing game." Result: It writes a simple script. But it might just pick a random number and ask once.

Round 2 (The Refinement):

"Make it so the user has 5 attempts. Tell them if they are 'Higher' or 'Lower'." Result: Now it has a loop and logic.

Round 3 (The Polish):

"Add error handling. If the user types 'banana', don't crash. Just say 'Please enter a number'." Result: A robust, beginner-friendly game.

The Final Code:

python
import random

def play_game():
    secret_number = random.randint(1, 100)
    attempts = 5

    print(f"I am thinking of a number between 1 and 100. You have {attempts} attempts.")

    for i in range(attempts):
        user_input = input("Guess: ")
        
        if not user_input.isdigit():
            print("Please enter a valid number.")
            continue
            
        guess = int(user_input)
        
        if guess == secret_number:
            print("You won! 🎉")
            return
        elif guess < secret_number:
            print("Higher! ⬆️")
        else:
            print("Lower! ⬇️")

    print(f"Game over! The number was {secret_number}.")

play_game()

Lesson: Don't expect perfection in one prompt. Build it layer by layer.

Cheat Sheet: Prompting Patterns 📜 ​

PatternGoalExample Prompt
PersonaBetter quality explanations."Act as a Senior Python Teacher..."
ELI5Understanding complex topics."Explain Recursion like I'm 5..."
SkeletonGetting started."Write the skeleton structure for this script..."
Logic TranslatorWriting complex if statements."Write code based on this English logic..."
Style CheckFixing messy code."Review this for PEP-8 and variable names..."
Data FormatterCleaning up text."Convert this raw text into a list of dicts..."

The "Temperature" Setting 🌡️ ​

Most AI tools have a "Temperature" setting (0.0 to 1.0).

  • Low (0.0 - 0.2): "The Robot". Logical, deterministic, good for code.
  • High (0.7 - 1.0): "The Artist". Creative, random, good for brainstorming ideas.

Tip: When asking for Code, use Low temperature (or tell the AI "Be precise and deterministic"). When asking for App Ideas, use High temperature.

Common Traps ⚠️ ​

  • Hallucinations: The AI might invent libraries that don't exist. Always verify imports!
  • Security: Never paste API keys, passwords, or sensitive company data into a public LLM.
  • Over-reliance: Don't copy-paste code you don't understand. If you can't explain it, don't ship it.