System vs User Role Distinction

Two Roles, Two Purposes

System Message = Developer-Defined Behavior

import anthropic

client = anthropic.Anthropic(api_key='YOUR_API_KEY')

response = client.messages.create(
    model='claude-opus-4-5',
    max_tokens=500,
    system='You are a customer support agent for TechCorp. '
           'Only answer questions about TechCorp products. '
           'Never discuss competitor products. '
           'Always be professional and concise.',
    messages=[
        {'role': 'user', 'content': 'What is your best laptop?'}
    ]
)

print(response.content[0].text)

User Message = Runtime Input

import anthropic

client = anthropic.Anthropic(api_key='YOUR_API_KEY')

def chat(system_prompt, conversation_history, user_input):
    conversation_history.append({'role': 'user', 'content': user_input})

    response = client.messages.create(
        model='claude-opus-4-5',
        max_tokens=500,
        system=system_prompt,  # Fixed: developer-controlled
        messages=conversation_history  # Dynamic: grows with each turn
    )

    assistant_reply = response.content[0].text
    conversation_history.append({'role': 'assistant', 'content': assistant_reply})
    return assistant_reply

history = []
sys = 'You are a helpful coding assistant. Only answer programming questions.'
print(chat(sys, history, 'How do I reverse a list in Python?'))

How Models Weight System vs User

Why System Prompts Persist Across Turns

import anthropic

client = anthropic.Anthropic(api_key='YOUR_API_KEY')

SYSTEM = 'You are a French language tutor. Respond in English but always include the French translation of key terms.'

history = []

def tutor(user_msg):
    history.append({'role': 'user', 'content': user_msg})
    r = client.messages.create(
        model='claude-opus-4-5',
        max_tokens=300,
        system=SYSTEM,   # Sent with every call — behavior persists
        messages=history
    )
    reply = r.content[0].text
    history.append({'role': 'assistant', 'content': reply})
    return reply

print(tutor('What is a verb?'))
print(tutor('Give me an example sentence.'))  # System rules still apply

System Prompt vs First User Message

The assistant Role in Multi-Turn Conversations

# Multi-turn conversation structure
messages = [
    {'role': 'user', 'content': 'What is machine learning?'},
    {'role': 'assistant', 'content': 'Machine learning is a type of AI that learns patterns from data...'},
    {'role': 'user', 'content': 'Can you give me a Python example?'},
    # Next call will add an assistant response here
]

# The model uses all prior turns as context,
# but always within the frame set by the system message.
print('Message history structure shown.')

Practical Role Separation in Applications

class ChatSession:
    def __init__(self, system_prompt):
        self.system = system_prompt  # Developer-controlled
        self.history = []

    def send(self, user_input):
        # Never mix user input into the system prompt
        # Never put behavioral rules into user messages
        self.history.append({'role': 'user', 'content': user_input})

        r = client.messages.create(
            model='claude-opus-4-5',
            max_tokens=500,
            system=self.system,
            messages=self.history
        )

        reply = r.content[0].text
        self.history.append({'role': 'assistant', 'content': reply})
        return reply

    def reset(self):
        self.history = []  # Clear turns but keep system prompt

session = ChatSession(system_prompt='You are a Python coding assistant.')
print(session.send('How do I read a file?'))

When to Use the System Message

OpenAI vs Anthropic System Message API

# OpenAI: system is a role in the messages array
import openai
client_oai = openai.OpenAI(api_key='YOUR_OPENAI_KEY')
response = client_oai.chat.completions.create(
    model='gpt-4o',
    messages=[
        {'role': 'system', 'content': 'You are a helpful assistant.'},
        {'role': 'user', 'content': 'Hello'}
    ]
)

# Anthropic: system is a top-level parameter
import anthropic
client_anth = anthropic.Anthropic(api_key='YOUR_ANTHROPIC_KEY')
response = client_anth.messages.create(
    model='claude-opus-4-5',
    max_tokens=100,
    system='You are a helpful assistant.',   # Top-level param
    messages=[{'role': 'user', 'content': 'Hello'}]
)

print('Both APIs support system prompts, different parameter structure.')

System Prompt Confidentiality

Quick Check

System vs User Role — Key Takeaways