What is the Gemini API's system instruction and how does it differ from a user prompt?

from google import genai
from google.genai import types

client = genai.Client()

# generateContent with system instruction:
response = client.models.generate_content(
    model="gemini-3.5-flash",
    contents="Explain recursion.",   # user input
    config=types.GenerateContentConfig(
        system_instruction="You are a Python tutor for beginners.
        Always give examples in Python.
        Keep explanations under 200 words.
        Never use jargon without explaining it first.",
    )
)
print(response.text)

# Interactions API with system instruction:
interaction = client.interactions.create(
    model="gemini-3.5-flash",
    input="Explain recursion.",
    system_instruction="You are a Python tutor for beginners...",
)

# Effective system instruction principles:
EFFECTIVE_SYSTEM_INSTRUCTION = """
Role: You are a senior Python code reviewer at a fintech company.

Constraints:
- Always check for security vulnerabilities first
- Flag any use of eval() or exec() as critical risk
- Use Decimal for financial calculations, never float
- Output in this format: Risk Level, Issues, Recommendations

Tone: Professional, direct, constructive
"""

# System instruction vs user prompt:
# System instruction: stable per session; shapes all responses; cached efficiently
# User prompt:        changes per turn; the actual task or question