From never having written a line in your life to directing whole projects. AI is the best programming teacher you've ever had — here's the map, step by step and at your own pace.
Five stages from zero to directing agents. Each one builds on the last, so take your time.
Curious
You haven't coded anything yet, and that's fine. Open a chat (ChatGPT, Claude or Gemini) and ask it to explain what a program is, which language suits you and why. Ask for tiny examples and have it explain them line by line.
Beginner
Use a cloud editor like Replit (no installing anything) and get a small program working. The golden rule: don't copy blindly. Paste every error into the chat and ask it to explain the cause, not just the fix.
Builder
Turn on GitHub Copilot's autocomplete in your editor: AI suggests as you type and you decide what to accept. You start reading code fluently and building projects with several moving parts.
Pilot
Move to an editor built for AI, like Cursor: describe what you want in plain language and the AI edits several files at once. Your job shifts from typing to directing, reviewing and testing.
Director
With agents like Claude Code you delegate whole tasks: the agent plans, writes, runs and fixes. You set the goal, review the result and keep the judgment. The key skill is no longer typing — it's knowing how to ask and verify.
There's no perfect language, only a first one. These are the most recommended to get going; what matters is picking one and starting.
The best first language for almost everyone: it reads almost like English and forgives beginner mistakes. It's the language of AI and data science.
Best for: Starting from zero, data and AI
Not programming, but the structure and style of web pages. The first step to building something visible and sharing it with the world in hours.
Best for: Your first web page
The language that brings the web to life: buttons, animations, whole apps inside the browser. If you want to make interactive things, start here.
Best for: Interactive sites and apps
The language for talking to databases: asking for, filtering and sorting information. Small, very useful and quick to learn.
Best for: Working with data
JavaScript with a safety net: it flags many errors before you run the code. The standard on serious projects, but best once you've got the basics down.
Best for: Large, team projects
The ecosystem moves fast, but these categories mark the path. Pick based on your level and what you want to do.
Chat assistants
Your private tutor: they explain concepts, generate examples, translate errors into human language and review your code. The starting point for everything.
Autocomplete
Suggests the next line or function right inside your editor as you type. Like a co-pilot that finishes the sentence you were about to write.
AI editor
A code editor built around AI: ask for changes in plain language and it applies them across the whole project, not just one line.
Coding agent
An agent that lives in your terminal: hand it a task and it plans, writes, runs and fixes itself until it's done. You direct.
App builders
Describe the app or website you want and they build it, working, in minutes. Great for prototyping and seeing results from day one.
Cloud editor
Code from your browser, nothing to install, with AI built in. Perfect for your first steps and for coding from any device.
AI editor
An AI editor with an agent called Cascade that understands your whole project and applies coordinated changes across several files at once.
Agentic IDE
Google's agent-first platform (powered by Gemini): agents plan and work at once across the editor, the terminal and the browser.
Coding agent
OpenAI's agent: delegate tasks and it works in your terminal or in the cloud, integrated with ChatGPT — even several at once.
Terminal agent
Google's terminal agent, open-source and free to start: it brings Gemini's power to your command line.
Open-source agent
An open coding agent that lives in the terminal and works with whatever model you choose, with no lock-in to a single company.
Terminal pair
Open-source pair programming from the terminal: it edits your repository and commits every change to Git automatically.
Four formulas that solve 80% of the daily grind. Copy them, adapt them, keep them.
I have this error: [paste the full error]. It's caused by this code: [paste the code]. Explain in plain language what it means, why it happens and how I fix it. Don't rewrite everything — just show me the line that changes.
Why it works: You ask for cause + minimal fix: you learn instead of pasting a patch you don't understand.
Explain what this code does step by step, as if I were a beginner. Point out the key parts and tell me what would happen if I deleted each one: [paste the code].
Why it works: Turns any snippet into a private lesson tailored to you.
Improve this code while keeping exactly the same behavior. Prioritize readability and clear names, and explain every important change: [paste the code].
Why it works: The “keep the behavior” keeps it from quietly changing your logic.
Write tests for this function covering the normal case, the edge cases and the errors. Explain what each one checks: [paste the function].
Why it works: Tests are your safety net: ask for them early and you'll code with confidence.
Small habits that separate depending on AI from knowing how to direct it.
The best time to write your first code was yesterday. The second best is now.
Practice with a template