Vicky Zhao
Free 5-day email course

Hand your first real task to Claude Code

A free 5-day course for analytical people who want AI to do actual work, not just answer questions.

In about 15 minutes a day, you'll pick one task from your real week, set up Claude Code safely, get a first pass, improve it, and learn how to verify what comes back. No coding experience required.

You might recognize this

If any of these sound like you, this is built for you.

Who this is for

You have the judgment. You're missing the teammate.

Most AI courses are written for engineers (you bounce off the jargon) or for absolute beginners (you waste your time on the basics). This one is for the people in between: founders, operators, PMs, writers, and analytical professionals. You can brief a sharp report. You've just never been shown how to point that skill at an AI agent.

15 mins/day

Small, daily, doable

One concept and one concrete action each day. No marathon sessions.

Your real work

Not toy demos

You pick what to hand Claude Code on Day 0, and you develop it around your work all week.

A teammate

Not a chatbot

By Day 5 you've handed real work to Claude Code and learned to verify what it gives back.

What you walk away with

From "AI should help somewhere" to one real handoff.

You do not need to learn to code. You need to learn to delegate to an agent: choose the right work, give it a safe folder, get a first pass, improve it, and check the result before you trust it.

Day 1

A real task

One specific piece of work from your actual week, chosen because it has leverage.

Day 2

A safe workspace

A clean Claude Code folder, with no sensitive files and no accidental whole-computer chaos.

Day 3

A first pass

A context prompt that gets Claude Code to inspect the folder and actually try the task.

Day 4

A better version

A way to tell Claude Code what's not good enough, get interviewed, and improve the same output.

Day 5

A second opinion

A sub-agent review pattern so you can check the improved work before you use it.

The plan

Your 5 days of training (Day 0 to Day 5)

Each day below is a preview. The full lesson and walkthrough video land in your inbox when you join. The arc is simple: find the right work, set up safely, create a first pass, make it better, then check it before you trust it.

Days 0–2

Find your workflow and get set up

Find the work worth handing over, learn why agents are different from chat, and get Claude Code running safely.

0

Day 0

Find your first handoff

Stop guessing what to use AI for. Run a 5-minute audit of your week and pick the one workflow worth handing to Claude Code first.

What you'll have: Your top 3 handoff candidates ranked, plus the one workflow you'll use for the course.
Walkthrough: the audit + picking your one taskUnlocks by email · ~5 min
Join to unlock →
1

Day 1

See why agents are different

Chat is for getting back and forth answers. An agent is for doing work. Once you understand the LLM and the agent harness, Claude Code starts to make sense.

What you'll have: A plain-English mental model for why agents can work across your files, plus a clearer picture of what your Day 0 task would need if you delegated it to an assistant.
Walkthrough: chat vs agent, and what a harness isUnlocks by email · ~6 min
Join to unlock →
2

Day 2

Set up the safe workspace

Install Claude Code, sign in, and give it one clean folder to work in. Not your whole computer. Not your sensitive files. One safe desk for your new agent.

What you'll have: Claude Code installed, pointed at a safe folder, and ready to touch only the files you intentionally give it.
Walkthrough: install + your first runUnlocks by email · ~8 min
Join to unlock →
Days 3–5

Delegate, then verify

Use your Day 0 task and Day 2 folder to create a first pass, improve it, then use a second agent to catch weak spots.

3

Day 3

Get Claude Code to do a first pass

Take the task you chose on Day 0, add the relevant files to your Day 2 folder, and give Claude Code enough context to try the work.

What you'll have: A first pass on your actual task, created from the same folder and context you'll keep using for the rest of the course.
Walkthrough: context + first passUnlocks by email · ~6 min
Join to unlock →
4

Day 4

Make the first pass better

Go back to the same Claude Code chat, tell it what is not good enough, let it interview you, then use expert frameworks to revise the output.

What you'll have: A revised version of the Day 3 output, improved with your critique and the frameworks that actually fit.
Walkthrough: interview + frameworks to improve the first passUnlocks by email · ~6 min
Join to unlock →
5

Day 5

Catch weak work before you use it

Don't trust the improved version just because it sounds confident. Use a sub-agent to evaluate what's strong, what's standard, and where the work might break.

What you'll have: A second-opinion loop for the improved Day 4 output, especially when the stakes are high or you're not the expert.
Walkthrough: how to judge AI work with a sub-agentUnlocks by email · ~7 min
Join to unlock →

Start with one real task.

Commit to 5 days, plus a Day 0 to find your starting point. Each day gives you one short lesson, one walkthrough video, and one concrete move on the same task: choose it, set up the folder, create a first pass, improve it, then review it.

Questions

Before you start

Do I need to know how to code?

No. The name is misleading. Claude Code is a tool for handing repeatable work to an agent, and this course is written for people who have never touched code. If you can brief an assistant, you can do this.

How much time does it take?

About 15 mins a day, Day 0 through Day 5. One concept, one concrete action. No marathon sessions.

What if I'm not technical?

That's exactly who this is for. Claude Code isn't programming. It's giving a sharp teammate clear briefs. If you can onboard a new hire or brief a contractor, you can do this.

Is it safe to use on my files?

You start in one clean folder, not your whole computer. Day 2 walks you through a safe workspace with nothing sensitive inside, and Day 5 teaches a verification loop before you trust the output.

Do I need a paid plan to use Claude Code?

Yes. But if you would rather not pay, you can run the same workflow on OpenAI's Codex, which has a limited free tier. This way you can test out using an agent before committing. Or if you have a subscription with another LLM, you can use their agent as well. The core idea is the same.

What do I actually get?

One email a day with a short written lesson and a walkthrough video, built around the real work you pick on Day 0. By the end you've selected one task, set up Claude Code safely, created a first pass, improved it with better context and frameworks, and learned to verify what comes back.

What happens after the 5 days?

You keep everything you built. If you want to go further, I run a live cohort called Claude Code Week, where we turn that first handoff into a working system around your real, messy work. I'll mention it near the end. No pressure either way.