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.

RAG pipeline — Ai Mastery

Day 7 · ~11 min●

RAG — Retrieval-Augmented Generation. Three steps:

Retrieve — top-k chunks by similarity to the query
Stuff — paste those chunks into the prompt as context
Generate — ask the LLM to answer using only the context

The model becomes a reading comprehension engine over your data, not a fact source.

python

contexts = [store[cid]["text"] for cid, _ in top_k(query, store, k=2)]
context_block = "\n".join(contexts)
prompt = f"""Use the context below to answer the question. If the answer isn't in the context, say so.

Context:
{context_block}

Question: {query}
Answer:"""

result = Agent(model).run_sync(prompt)

And the LLM only knows what's in the context block?

Right — that's the whole point. You're not asking what does the model know. You're asking what does this paragraph say. The model becomes grounded.

What if my retrieval misses?

Then the answer is wrong, and the LLM may hallucinate to fill the gap. Tomorrow's lesson — citations — makes the failure visible. The lesson after — failure modes — categorises what goes wrong.

RAG pipeline — retrieve, stuff, answer

A single function that ties the week together:

python

def rag_answer(query, store, k=2):
    # 1. Retrieve
    top = top_k(query, store, k)
    contexts = [store[cid]["text"] for cid, _ in top]
    context_block = "\n".join(contexts)

    # 2. Stuff
    prompt = f'''Use the context to answer the question. If unsure, say "I don't know".

Context:
{context_block}

Question: {query}
Answer:'''

    # 3. Generate
    return Agent(model).run_sync(prompt).output

What changes vs a plain LLM call

Without RAG: model answers from training data. Stale, possibly wrong, no source.

With RAG: model answers from your data. Fresh, traceable, the source is in the prompt.

Cost shape

Embedding the query: 1 call (cheap embedding API)
Cosine over store: pure Python, free
LLM completion: 1 call, prompt tokens = system + context + query, completion tokens = answer

The LLM call is the dominant cost. Smaller k = fewer prompt tokens = cheaper.

What can go wrong (week 2 covers each)

Hallucination — model invents facts when retrieval misses (L9)
Out-of-corpus — query is about something not in your data (L9)
Stale answer — corpus is outdated; model agrees confidently (need to refresh embeddings)
Untraceable — answer cites no source, you can't audit (L8 — citations)

For today's lesson the corpus is small enough that retrieval is rarely wrong. The point is the pipeline — retrieve → stuff → answer — wired end to end.

Day 7 · ~11 min●

RAG — Retrieval-Augmented Generation. Three steps:

Retrieve — top-k chunks by similarity to the query
Stuff — paste those chunks into the prompt as context
Generate — ask the LLM to answer using only the context

The model becomes a reading comprehension engine over your data, not a fact source.

python

contexts = [store[cid]["text"] for cid, _ in top_k(query, store, k=2)]
context_block = "\n".join(contexts)
prompt = f"""Use the context below to answer the question. If the answer isn't in the context, say so.

Context:
{context_block}

Question: {query}
Answer:"""

result = Agent(model).run_sync(prompt)

And the LLM only knows what's in the context block?

Right — that's the whole point. You're not asking what does the model know. You're asking what does this paragraph say. The model becomes grounded.

What if my retrieval misses?

RAG pipeline — retrieve, stuff, answer

A single function that ties the week together:

python

def rag_answer(query, store, k=2):
    # 1. Retrieve
    top = top_k(query, store, k)
    contexts = [store[cid]["text"] for cid, _ in top]
    context_block = "\n".join(contexts)

    # 2. Stuff
    prompt = f'''Use the context to answer the question. If unsure, say "I don't know".

Context:
{context_block}

Question: {query}
Answer:'''

    # 3. Generate
    return Agent(model).run_sync(prompt).output

What changes vs a plain LLM call

Without RAG: model answers from training data. Stale, possibly wrong, no source.

With RAG: model answers from your data. Fresh, traceable, the source is in the prompt.

Cost shape

Embedding the query: 1 call (cheap embedding API)
Cosine over store: pure Python, free
LLM completion: 1 call, prompt tokens = system + context + query, completion tokens = answer

The LLM call is the dominant cost. Smaller k = fewer prompt tokens = cheaper.

What can go wrong (week 2 covers each)

Hallucination — model invents facts when retrieval misses (L9)
Out-of-corpus — query is about something not in your data (L9)
Stale answer — corpus is outdated; model agrees confidently (need to refresh embeddings)
Untraceable — answer cites no source, you can't audit (L8 — citations)

For today's lesson the corpus is small enough that retrieval is rarely wrong. The point is the pipeline — retrieve → stuff → answer — wired end to end.

RAG pipeline — retrieve, stuff, answer

What changes vs a plain LLM call

Cost shape

What can go wrong (week 2 covers each)

RAG pipeline — retrieve, stuff, answer

What changes vs a plain LLM call

Cost shape

What can go wrong (week 2 covers each)

RAG pipeline — retrieve, stuff, answer

What changes vs a plain LLM call

Cost shape

What can go wrong (week 2 covers each)

Sign up to practice

RAG pipeline — retrieve, stuff, answer

What changes vs a plain LLM call

Cost shape

What can go wrong (week 2 covers each)

Sign up to practice