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.

Rank: Agent Picks the Most Informative URL — Ai For Makers

Day 6 · ~12 min●

You have a list of five URLs with titles and you want the agent to pick the most informative one. How do you make sure it returns a URL from the list, not a paraphrase?

Add a system_prompt saying "reply only with one of these URLs"? That still lets the model invent something.

A prompt is a hint. A type is a contract. Build Literal[url1, url2, ...] dynamically from the input list and pass it as result_type. Now the only valid replies are the exact URLs you handed in:

python

from typing import Literal
urls = [r["url"] for r in results]
agent = Agent(model, result_type=Literal[tuple(urls)])

The type is generated from the data? I pass the exact URL tuple as the Literal members so the agent cannot drift?

Exactly — a runtime-built type. The agent sees the schema, picks one, and PydanticAI validates the answer. Wrapped with a short ranking prompt:

python

def select_best_url(results: list) -> str:
    urls = [r["url"] for r in results]
    agent = Agent(
        model,
        result_type=Literal[tuple(urls)],
        system_prompt="Pick the single most informative URL from the list."
    )
    listing = "\n".join(f"{r['title']}: {r['url']}" for r in results)
    result = agent.run_sync(listing)
    print(f"Agent pick: {result.output}")
    return result.output

And if someone passes a list of one URL? The Literal still validates, right?

It does — a one-member Literal is still valid. For larger lists you feed every title plus URL into the prompt so the model has context to choose, but the output type guarantees only a URL from the input comes out. Invention is impossible.

So this pattern fits any ranking task — best candidate, highest-priority ticket, top reply — whenever the answer is one of a known set?

That is the generalization. Any selection over a finite input set is a Literal type built from the set itself. Typed ranking, no string matching, no fallback logic.

Runtime-Built Literal Types

TL;DR: build Literal[tuple(options)] from your input so the agent must pick a real member.

Literal[tuple(urls)] — closed set derived from the call arguments
No invention possible — the validator rejects anything not in the tuple
system_prompt — guides selection criteria ("most informative", "highest priority")

When to build Literal at runtime

Set known at code time	Set depends on input
`Literal["a","b","c"]`	`Literal[tuple(options)]`
Static enum	Dynamic ranking

A Literal built from the input is the simplest safe way to ask the agent to select.

Day 6 · ~12 min●

You have a list of five URLs with titles and you want the agent to pick the most informative one. How do you make sure it returns a URL from the list, not a paraphrase?

Add a system_prompt saying "reply only with one of these URLs"? That still lets the model invent something.

A prompt is a hint. A type is a contract. Build Literal[url1, url2, ...] dynamically from the input list and pass it as result_type. Now the only valid replies are the exact URLs you handed in:

python

from typing import Literal
urls = [r["url"] for r in results]
agent = Agent(model, result_type=Literal[tuple(urls)])

The type is generated from the data? I pass the exact URL tuple as the Literal members so the agent cannot drift?

Exactly — a runtime-built type. The agent sees the schema, picks one, and PydanticAI validates the answer. Wrapped with a short ranking prompt:

python

def select_best_url(results: list) -> str:
    urls = [r["url"] for r in results]
    agent = Agent(
        model,
        result_type=Literal[tuple(urls)],
        system_prompt="Pick the single most informative URL from the list."
    )
    listing = "\n".join(f"{r['title']}: {r['url']}" for r in results)
    result = agent.run_sync(listing)
    print(f"Agent pick: {result.output}")
    return result.output

And if someone passes a list of one URL? The Literal still validates, right?

So this pattern fits any ranking task — best candidate, highest-priority ticket, top reply — whenever the answer is one of a known set?

That is the generalization. Any selection over a finite input set is a Literal type built from the set itself. Typed ranking, no string matching, no fallback logic.

Runtime-Built Literal Types

TL;DR: build Literal[tuple(options)] from your input so the agent must pick a real member.

Literal[tuple(urls)] — closed set derived from the call arguments
No invention possible — the validator rejects anything not in the tuple
system_prompt — guides selection criteria ("most informative", "highest priority")

When to build Literal at runtime

Set known at code time	Set depends on input
`Literal["a","b","c"]`	`Literal[tuple(options)]`
Static enum	Dynamic ranking

A Literal built from the input is the simplest safe way to ask the agent to select.

def select_best_url(results: list) -> str: urls = [r["url"] for r in results] agent = Agent( model, result_type=Literal[tuple(urls)], system_prompt="Pick the single most informative URL from the list." ) listing = "\n".join(f"{r['title']}: {r['url']}" for r in results) result = agent.run_sync(listing) print(f"Agent pick: {result.output}") return result.output

Runtime-Built Literal Types

TL;DR: build Literal[tuple(options)] from your input so the agent must pick a real member.

Literal[tuple(urls)] — closed set derived from the call arguments

No invention possible — the validator rejects anything not in the tuple

system_prompt — guides selection criteria ("most informative", "highest priority")

When to build Literal at runtime

Set known at code time	Set depends on input
`Literal["a","b","c"]`	`Literal[tuple(options)]`
Static enum	Dynamic ranking

A Literal built from the input is the simplest safe way to ask the agent to select.

Runtime-Built Literal Types

TL;DR: build Literal[tuple(options)] from your input so the agent must pick a real member.

Literal[tuple(urls)] — closed set derived from the call arguments

No invention possible — the validator rejects anything not in the tuple

system_prompt — guides selection criteria ("most informative", "highest priority")

When to build Literal at runtime

Set known at code time	Set depends on input
`Literal["a","b","c"]`	`Literal[tuple(options)]`
Static enum	Dynamic ranking

A Literal built from the input is the simplest safe way to ask the agent to select.

Runtime-Built Literal Types

When to build Literal at runtime

Runtime-Built Literal Types

When to build Literal at runtime

Runtime-Built Literal Types

When to build Literal at runtime

Sign up to practice

Runtime-Built Literal Types

When to build Literal at runtime

Sign up to practice