AI Programming for Complete Beginners: A Full Guide from Installation to Real Projects

AI Programming Isn't Just for Programmers

Many people shy away the moment they hear "programming," assuming it's something only professional developers can do. But as AI programming tools have matured, anyone can now accomplish tasks ranging from writing programs to managing files simply by having a natural language conversation with AI. This article, based on a systematic AI course by Bilibili creator Datong, outlines a complete beginner's path into AI programming.

The core idea is simple: You don't need to know how to write code — you just need to know how to describe what you want. The author is not a programmer; all their understanding of code was accumulated through using AI tools. They've even independently built websites and published apps on the Apple App Store.

What Is AI Programming? The Leap from Chat to Action

Web Chat vs. Local Operations

When we use tools like Doubao or ChatGPT, they can only give us text-based suggestions on a webpage. Even if they provide code, you still need to manually copy it to your computer and run it yourself. AI programming tools (like Claude Code), once installed on your computer, can:

Directly read files on your computer
Create and modify files for you
Run code
Adjust system settings (such as modifying proxy configurations)

In short, AI used to only "talk" — now it has "hands and feet" and can directly execute operations on your computer. This transformation from "advisor" to "executor" is the fundamental difference between AI programming tools and regular chatbots — they have read/write access to your local file system and can act like an assistant sitting at your computer, actually doing the work.

Comparison of Major AI Programming Tools

There are many AI programming tools on the market today, including:

Claude Code: Made by Anthropic, command-line based, great for beginners
Codex: Made by OpenAI
Cursor: An early popular AI programming IDE
Gemini CLI (Anti-Gravity): Made by Google, with a standalone client interface
TRAE: Made by ByteDance, with positive user feedback
Open Claude (Lobster): A higher-level tool with advanced permissions like scheduled tasks

Recommendation: You don't need to learn all of them — just pick one. Start with Claude Code, then try Gemini CLI or TRAE later. The core difference between these tools isn't the software itself, but the AI brain (model) they connect to — whether it's GPT, Gemini, or Claude, the model determines the capability ceiling.

Here's a key concept to understand: AI programming tools are essentially a combination of a "shell" plus a "brain." The shell (like Claude Code) handles interaction with your computer, file management, and command execution; the brain (like Claude, GPT-4, Gemini) understands your requirements and generates solutions. The same shell can connect to different brains, just like the same computer can run different operating systems.

Complete Claude Code Installation Guide

Step 1: Install the Node.js Environment

Before installing Claude Code, you need to set up the Node.js environment. Node.js is a JavaScript runtime environment built on Chrome's V8 engine that allows JavaScript to run beyond the browser — on servers and local computers. npm (Node Package Manager) is the package management tool bundled with Node.js, similar to an app store on your phone, allowing developers to install various open-source tools and libraries with a single command. Claude Code requires Node.js because it's distributed as a command-line tool package via npm. For non-programmers, just think of Node.js as "the infrastructure that lets AI programming tools run on your computer."

Installation steps:

Download the Node.js Windows installer
Double-click to install, clicking "Next" all the way through
Verify installation: Press Win+R, type cmd, then enter node -v in the command line — if a version number appears, installation was successful

Step 2: Install Claude Code

The command line (also called terminal, Terminal, or CMD) is a way to operate your computer by typing text commands, in contrast to the graphical interface we're used to (clicking icons, dragging files). Although it might feel like "going back to the DOS era," the command line is extremely efficient in development because it can precisely execute complex operations, batch-process tasks, and automate workflows. Claude Code uses the command line as its interface because it needs direct access to the file system and program execution — the command line provides the most direct permission channel. For beginners, just remember: open the command line window, type a command, press Enter to execute — that's all you need.

Global installation (recommended): Enter in the command line:

npm install -g @anthropic-ai/claude-code

Local installation: Run the same command without the -g flag in your desired folder's command line. The advantage of local installation is that deleting the folder completely uninstalls it.

After installation, type claude (global) or npx claude (local) in the command line to launch Claude Code.

Step 3: Configure the Model and API Key

Claude Code needs to connect to an AI model to function. Here's an important concept: an API Key (Application Programming Interface Key) is essentially a "digital key" that verifies your identity and authorizes access to AI model services. When you enter a prompt in Claude Code, the tool sends your request over the internet to the AI model on a remote server, which processes it and returns results. A token is the basic unit of measurement for how large language models process text — Chinese text consumes roughly 1 token per 1-2 characters, and the model consumes tokens every time it reads your context and generates a response. This is why pay-per-use billing can be extremely expensive — a complex project conversation can easily consume hundreds of thousands of tokens.

Here's an important tip: Don't choose pay-per-use billing, because token consumption is staggering. Two cost-effective subscription plans are recommended:

Alibaba Bailian: Basic plan around ¥40/month (¥7.9 for new users), includes models like Qianwen 3.5 Plus, GLM5, Kimi, with unlimited usage
Volcano Engine (Ark Coding Plan): Also around ¥40/month, with detailed Claude Code integration guides

Configuration steps:

Obtain your API Key from the subscription platform
Create a configuration file via command line, filling in the API address, Key, and model name
Restart Claude Code to begin using it

The author has also prepared a one-click installation script that automatically completes global installation and configuration with a double-click — perfect for beginners.

Hands-On Demo: Build a Memo App in 5 Minutes

The Right Approach

The core workflow for building projects with Claude Code:

Create a new folder: Every project should have its own working directory
Discuss before executing: Don't ask AI to start building immediately — first have it propose a plan
Confirm before starting: Analyze technical choices and feature planning from a professional perspective

Example prompt:

"I want to make a front-end web memo app with a nice-looking interface. I don't know programming — give me suggestions. Propose a plan first, and I'll confirm before you start."

The AI then provided feature suggestions for basic, standard, and full versions, along with a technical plan using the Vue framework with a modern dark theme. The "front-end" mentioned here refers to the interface part that users directly see and interact with, including webpage layout, colors, buttons, and animations, primarily using HTML, CSS, and JavaScript. Vue is one of the most popular front-end frameworks, providing a set of code organization standards and ready-made component libraries that significantly boost development efficiency. For AI programming users, understanding these concepts matters because: when you tell AI "make a front-end webpage," it knows to only handle the presentation layer; if you say "I need user login and data saving," it will also build a back-end service (the part running on a server responsible for data storage and business logic).

After confirmation, the AI automatically began writing code, creating files, and starting the service.

How to Handle Errors

During the demo, a "cannot access website" error occurred — the cause was a port conflict. A port is a logical concept in computer networking, similar to a room number in a building. When a web application starts, it occupies a specific port on your computer (like 3000, 8080, etc.) to provide its service. If that port is already occupied by another program, the new program can't start — just like trying to enter a room that's already occupied. This is one of the most common issues in development.

The solution is very simple: Just tell the AI what problem you encountered, and it will automatically diagnose and fix it (such as closing the program occupying the port or switching to a different port number).

This also highlights differences in model capabilities: domestic models (like GLM5) handle simple tasks adequately, but their efficiency noticeably falls behind Gemini or Claude when facing complex problems.

Advanced Case: Building a File Organizer with AI

The author shared a practical tool built with Gemini CLI in just 5 minutes — a file organization program:

Reads files from multiple drives (C/D/E)
Displays preview thumbnails and file icons
Supports drag-and-drop categorization
Right-click multi-select functionality

The entire process was completed through natural language conversation, from stating requirements to iterating on features, without manually writing a single line of code. This case perfectly demonstrates the practical value of AI programming tools — they can do more than just "toy projects" and can solve real pain points in daily work.

Key Advice for Complete Beginners

Describing Requirements Matters More Than Reading Code

The author explicitly states: "From the very beginning until now, being able to independently build websites and publish apps, I have zero experience modifying code myself." What you need to strengthen is:

User mindset: Being clear about what you actually want
Communication skills: Being able to describe requirements clearly and specifically
Basic knowledge: Understanding fundamental concepts like front-end/back-end (you can ask AI about these too)

How to Avoid Wasting Tokens

Poorly described requirements will send AI down many wrong paths, wasting tokens while producing poor results. Suggestions:

Learn basic technical concepts before stating requirements
Plan large projects first, then execute in stages
Small tasks can be requested directly, but clearly describe the expected outcome

Recommended Learning Path

Install Claude Code and start with domestic models (Alibaba Bailian/Volcano Engine plans)
Your first question to AI should be: "How can I use you more effectively?"
Start practicing with simple tools (memo apps, file organizers, etc.)
Gradually learn advanced concepts like Skills and MCP
Later, upgrade to Gemini or Claude for stronger programming capabilities

MCP (Model Context Protocol) is an open standard proposed by Anthropic that enables AI models to securely connect to external data sources and tools. Think of MCP as AI's "universal adapter" — through it, AI can not only read and write local files but also connect to databases, call web services, operate browsers, and even control smart home devices. The emergence of MCP means the capability boundaries of AI programming tools are rapidly expanding, moving from pure code generation toward a true "digital employee" role. Mastering MCP configuration and usage will dramatically expand what your AI assistant can do.

Conclusion

The barrier to AI programming has dropped to an unprecedented low. You don't need a programming background, you don't need English proficiency — all you need is a subscription plan and the ability to clearly express your requirements. From the moment you install Claude Code, you have a 24/7 all-capable assistant — it can write programs, organize files, solve technical problems, and even help you learn anything you want to learn.

From a broader perspective, the proliferation of AI programming tools is redefining what "technical ability" means. In the past, technical ability equaled the ability to write code; in the future, it will increasingly mean the ability to clearly express requirements, understand system architecture, and effectively coordinate AI tools. This isn't the end for programmers — it's an opportunity for everyone. When the barrier to programming disappears, what becomes truly scarce will be creativity, judgment, and problem-solving thinking.