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.

AI Patterns — How Far You've Come — Ai Patterns

Day 30 · ~12 min●

Thirty days. From add(7, 8) = 15 to a multi-tool agent grading itself across an eval suite with a weighted rubric. Same six prompts you saw on day 1 — rate yourself again.

What's next?

You have the patterns kit. Combined with Python Foundations (the language), Automation Foundations (acting on the world via Composio), and AI Foundations (calling LLMs), you can compose a script for any task you face — classify, extract, transform, route, decide, validate, iterate.

Beyond v1: AI Advanced would add embeddings, RAG (retrieval-augmented generation), model routing, response caching, production guardrails (PII detection, jailbreak resistance), and fine-tuning. All sit on top of the primitives you have now — they're refinements, not replacements.

What should I do this week?

Build something. Not a product — a small script for a real itch. Pick one task you do manually that has structure: triaging a list of items, drafting responses to a recurring kind of message, summarising a daily digest. Apply the patterns:

Define the eval set — 5 (input, expected) pairs from your real work
Write narrow tools — one purpose each
Wrap Composio actions you need
Build the agent + rubric
Iterate via evals — change one thing, re-score, ship when good enough

That's the move from "finished a track" to "can ship LLM features". The kit you have is enough.

What you can do now

24 small Python LLM-pattern scripts across four weeks plus two synthesis lessons. Your kit:

Capability	Lessons
Tool calling fundamentals	L1 (why), L2 (schema), L3 (single), L4 (multi-step)
Reasoning patterns	L5 (CoT), L6 (few-shot CoT)
Pipelines	L8 (chained), L9 (classify-branch)
Validation + critique	L10 (validation), L11 (self-critique), L12 (moderation), L13 (eval criteria)
Agents	L15 (concept), L16 (loop), L17 (multi-tool)
Evals	L18 (suite), L19 (iteration), L20 (synthesis)
Production patterns	L22 (typed), L23 (retry+), L24 (rubrics), L25 (planning chains), L26 (agent + Composio), L27 (integration synthesis)

Any LLM agent or pattern-using script you'll read or write is a composition of these.

What this track did NOT teach

Deferred to AI Advanced (when v1 expands):

Embeddings — semantic representations of text
RAG — retrieval-augmented generation, grounding in your data
Model routing — picking the right model per task (cost vs quality)
Response caching — short-circuiting repeated prompts
Production guardrails — PII detection, content moderation APIs, jailbreak resistance
Fine-tuning — model customisation

You don't need any of these for your first dozen useful AI scripts. You have enough.

The v1 picture

Nine tracks. Three layers:

Python (Foundations / Patterns / Mastery) — the language
Automation (Foundations / Patterns / Mastery) — acting on the world via Composio
AI (Foundations / Patterns / [Advanced — deferred]) — calling LLMs, then giving them tools, validation, and agency

After eight tracks you can compose. After all nine you can compose with the production primitives. The platform's promise is the kit, not the prebuilt product.

Your move

Pick a real itch. Apply the kit. The discipline is:

Eval first. Write the (input, expected) pairs before the prompt. The pairs ARE the spec.
Smallest tools. One purpose each. Easy to audit, easy for the model to use.
One change per iteration. Measure. Keep what helped.
Cap loops. Steps, retries, agent iterations — all capped, all loud on cap-hit.
Validate everything. Output validation isn't optional once anything other than your eyes is consuming the response.

Rate the prompts below as honestly as you did on day 1.

Day 30 · ~12 min●

Thirty days. From add(7, 8) = 15 to a multi-tool agent grading itself across an eval suite with a weighted rubric. Same six prompts you saw on day 1 — rate yourself again.

What's next?

What should I do this week?

Define the eval set — 5 (input, expected) pairs from your real work
Write narrow tools — one purpose each
Wrap Composio actions you need
Build the agent + rubric
Iterate via evals — change one thing, re-score, ship when good enough

That's the move from "finished a track" to "can ship LLM features". The kit you have is enough.

What you can do now

24 small Python LLM-pattern scripts across four weeks plus two synthesis lessons. Your kit:

Capability	Lessons
Tool calling fundamentals	L1 (why), L2 (schema), L3 (single), L4 (multi-step)
Reasoning patterns	L5 (CoT), L6 (few-shot CoT)
Pipelines	L8 (chained), L9 (classify-branch)
Validation + critique	L10 (validation), L11 (self-critique), L12 (moderation), L13 (eval criteria)
Agents	L15 (concept), L16 (loop), L17 (multi-tool)
Evals	L18 (suite), L19 (iteration), L20 (synthesis)
Production patterns	L22 (typed), L23 (retry+), L24 (rubrics), L25 (planning chains), L26 (agent + Composio), L27 (integration synthesis)

Any LLM agent or pattern-using script you'll read or write is a composition of these.

What this track did NOT teach

Deferred to AI Advanced (when v1 expands):

Embeddings — semantic representations of text
RAG — retrieval-augmented generation, grounding in your data
Model routing — picking the right model per task (cost vs quality)
Response caching — short-circuiting repeated prompts
Production guardrails — PII detection, content moderation APIs, jailbreak resistance
Fine-tuning — model customisation

You don't need any of these for your first dozen useful AI scripts. You have enough.

The v1 picture

Nine tracks. Three layers:

Python (Foundations / Patterns / Mastery) — the language
Automation (Foundations / Patterns / Mastery) — acting on the world via Composio
AI (Foundations / Patterns / [Advanced — deferred]) — calling LLMs, then giving them tools, validation, and agency

After eight tracks you can compose. After all nine you can compose with the production primitives. The platform's promise is the kit, not the prebuilt product.

Your move

Pick a real itch. Apply the kit. The discipline is:

Eval first. Write the (input, expected) pairs before the prompt. The pairs ARE the spec.
Smallest tools. One purpose each. Easy to audit, easy for the model to use.
One change per iteration. Measure. Keep what helped.
Cap loops. Steps, retries, agent iterations — all capped, all loud on cap-hit.
Validate everything. Output validation isn't optional once anything other than your eyes is consuming the response.

Rate the prompts below as honestly as you did on day 1.

What you can do now

What this track did NOT teach

The v1 picture

Your move

What you can do now

What this track did NOT teach

The v1 picture

Your move

What you can do now

What this track did NOT teach

The v1 picture

Your move

Sign up to practice

What you can do now

What this track did NOT teach

The v1 picture

Your move

Sign up to practice