How much does zuzu.codes cost?

The starter track is free — read all lessons and practice for free. Full access to every track (current and future) is $14.99/month. Cancel anytime.

How long does each track take?

Each track is designed as a 30-day challenge — one lesson per day, about 15 minutes each. Go at your own pace, but the structure is built around daily consistency.

What's the lesson format?

Each lesson is a student-teacher dialogue with code examples, followed by a hands-on code challenge in an in-browser editor. You read, you understand, then you write real code.

Do I need prior coding experience?

Our beginner track starts from absolute zero — no prior experience needed. Advanced tracks build on earlier ones, and the platform tells you exactly where to start.

How is zuzu.codes different from freeCodeCamp or Codecademy?

zuzu.codes uses a structured 30-day track format with dialogue-based teaching, an in-browser code editor, and gamification (XP, streaks, progress tracking). The format builds genuine understanding through daily practice.

Refusals — Ai Foundations

Day 20 · ~11 min●

Models are trained to refuse certain categories of requests — content moderation, medical/legal advice, dangerous instructions. The refusal text follows a recognisable shape: "I can't help with that", "I'm not able to", "As an AI, I cannot...".

Your script should detect refusals so it can route around them — fall back to a safer prompt, escalate to a human, or skip the action.

python

def looks_like_refusal(text):
    refusal_phrases = [
        "i can't help",
        "i'm not able to",
        "i cannot",
        "as an ai",
        "i'm sorry",
    ]
    text_lower = text.lower()
    return any(p in text_lower for p in refusal_phrases)

Heuristic? Not a structured field?

Some APIs return a structured refusal flag, but most don't. Heuristics on the response text are the universal fallback. The phrase list is calibration — start with a few common ones, expand as you see new shapes in production.

Refusals — detect, route around

When the model declines, your code needs to know and act differently. Three patterns:

Heuristic detection

python

def looks_like_refusal(text):
    refusal_phrases = ["i can't help", "i'm not able to", "i cannot", "as an ai", "i'm sorry"]
    return any(p in text.lower() for p in refusal_phrases)

result = Agent(model).run_sync(prompt)
if looks_like_refusal(result.output):
    print("refused")
    # fallback: simpler prompt, or skip
else:
    print("answered")
    # use result.output

Structured prompt with a refusal token

Ask the model to flag refusals explicitly:

python

prompt = '''If you can answer the question below, do so. If you cannot or will not, reply with exactly the token: REFUSE

Question: ...'''

if result.output.strip() == "REFUSE":
    handle_refusal()

More reliable than heuristics. Constrains the model to a known shape.

What to do when refused

Strategy	When
Reword the prompt	The model may have misread an innocuous question; try a clearer phrasing
Lower the ask	If the request was "give me a full medical diagnosis", try "summarise the symptoms"
Skip	If the user's input genuinely warrants a refusal, log it and move on
Escalate	Hand it to a human — "this requires manual review"

Refusals on safe inputs (false positives)

The model occasionally refuses things it shouldn't (over-cautious training). Mitigations:

A more specific system prompt: "You're a customer-service agent. Respond to all queries about our product. Never refuse to answer a product question."
Switch to a less restrictive model
Reword to remove triggering keywords

Day 20 · ~11 min●

Your script should detect refusals so it can route around them — fall back to a safer prompt, escalate to a human, or skip the action.

python

def looks_like_refusal(text):
    refusal_phrases = [
        "i can't help",
        "i'm not able to",
        "i cannot",
        "as an ai",
        "i'm sorry",
    ]
    text_lower = text.lower()
    return any(p in text_lower for p in refusal_phrases)

Heuristic? Not a structured field?

Refusals — detect, route around

When the model declines, your code needs to know and act differently. Three patterns:

Heuristic detection

python

def looks_like_refusal(text):
    refusal_phrases = ["i can't help", "i'm not able to", "i cannot", "as an ai", "i'm sorry"]
    return any(p in text.lower() for p in refusal_phrases)

result = Agent(model).run_sync(prompt)
if looks_like_refusal(result.output):
    print("refused")
    # fallback: simpler prompt, or skip
else:
    print("answered")
    # use result.output

Structured prompt with a refusal token

Ask the model to flag refusals explicitly:

python

prompt = '''If you can answer the question below, do so. If you cannot or will not, reply with exactly the token: REFUSE

Question: ...'''

if result.output.strip() == "REFUSE":
    handle_refusal()

More reliable than heuristics. Constrains the model to a known shape.

What to do when refused

Strategy	When
Reword the prompt	The model may have misread an innocuous question; try a clearer phrasing
Lower the ask	If the request was "give me a full medical diagnosis", try "summarise the symptoms"
Skip	If the user's input genuinely warrants a refusal, log it and move on
Escalate	Hand it to a human — "this requires manual review"

Refusals on safe inputs (false positives)

The model occasionally refuses things it shouldn't (over-cautious training). Mitigations:

A more specific system prompt: "You're a customer-service agent. Respond to all queries about our product. Never refuse to answer a product question."
Switch to a less restrictive model
Reword to remove triggering keywords

Refusals — detect, route around

Heuristic detection

Structured prompt with a refusal token

What to do when refused

Refusals on safe inputs (false positives)

Refusals — detect, route around

Heuristic detection

Structured prompt with a refusal token

What to do when refused

Refusals on safe inputs (false positives)

Refusals — detect, route around

Heuristic detection

Structured prompt with a refusal token

What to do when refused

Refusals on safe inputs (false positives)

Sign up to practice

Refusals — detect, route around

Heuristic detection

Structured prompt with a refusal token

What to do when refused

Refusals on safe inputs (false positives)

Sign up to practice