Claude Code 3-Day Hands-on: This AI Programmer Made Me Both Happy and Scared

Hello, I'm Wanfeng, a programmer who experiments with various AI projects.

This week, Anthropic released Claude Code, dubbed "the AI programmer that lives in your terminal." As a 10-year veteran coder, I tried it out the moment it was released. After 3 days of hands-on testing?

I'm both happy and scared.

I'm happy because the efficiency is really high. I'm scared because… this thing is progressing way too fast.

Day 1: First Encounter, Slightly Impressed

Installation was simple, one command:

1	npm install -g @anthropics/claude-code

Then run claude in your project directory to enter conversation mode.

My first test: let it understand my codebase.

Me: What does this project do?
Claude: This is a Python office automation library. The main features are...

It accurately identified the project's core features, module structure, and even pointed out which file is the entry point.

This is faster than me onboarding a new colleague to the project.

Day 2: Real-world Bug Fix, Solved in 5 Minutes

I happened to hit a production issue: a user reported an error when processing Excel.

Old me: read the log → locate the code → set breakpoints → debug → fix. At least half an hour.

With Claude Code:

Me: The user reported this error. What does it mean?
[Paste error message]

Claude: This is an openpyxl version compatibility issue. When the cell value is empty...
You should add a Z check in the Y function of file X.

Me: Help me fix it.
Claude: Done. Want me to run the tests?

The whole thing took 5 minutes, including writing test cases.

Day 3: Starting to Panic

On the third day, I asked it to implement a new feature: batch-process PDFs and generate a summary report.

After I briefly described the requirements, Claude Code:

Automatically installed the needed libraries (PyPDF2, pandas)
Wrote the core processing logic
Added exception handling
Generated test data
Ran tests to verify

I only said a few sentences, and it did what would have taken me 2 hours.

At that moment, I really started to panic: if AI can do all this, why should I still learn programming?

After Cooling Down: AI Is an Amplifier, Not a Replacement

After thinking for a long time, I realized one fact:

Claude Code is strong, but it needs someone to tell it what to do.

Just like a calculator is powerful, but you still need to know what to calculate.

The truly valuable skills have become:

Defining the problem (knowing what to solve)
Breaking down tasks (splitting a big goal into small steps)
Judging quality (knowing whether the AI did it right)
System design (controlling the overall architecture)

These abilities are exactly what learning programming can develop.

Advice for Beginners

If you haven't started learning programming yet, seeing Claude Code might make you think: "If AI can write code, why should I learn?"

My answer: Precisely because AI exists, learning programming is more important than ever.

Here are three reasons:

You can't ask good questions, AI can't give good answers.
If you don't understand programming, you won't even know how to describe the requirement.
You can't judge, and you'll be led astray by AI.
AI also makes mistakes; without a foundation, you won't notice.
You can't extend, so you can only do the simplest things.
Slightly complex requirements need to combine multiple skills.

The future programmer isn't being eliminated—they're being upgraded.

From "someone who writes code" to "someone who directs AI to write code."