Before you start: the programming you need

In one line: This guide teaches AI engineering from zero AI knowledge — but it does assume a little programming knowledge. This page lists exactly what, teaches each idea in plain English, and points you to a full course if you've never coded at all.

In plain English

Good news: the programming bar for AI engineering is low. You will not write algorithms or build compilers. You'll write short scripts that send text to an API and handle what comes back. If you can read the five ideas below and they make sense, you're ready for the whole guide. If they're brand new, spend a week with a beginner programming course first — it pays off immediately.

The five ideas

1. Variables and functions

A variable is a named box that holds a value. A function is a named, reusable block of code that takes inputs and returns an output.

price_per_token = 0.000003          # a variable

def cost(tokens):                   # a function
    return tokens * price_per_token

print(cost(1000))                   # 0.003

Trace it: cost(1000) runs the function body with tokens = 1000, multiplies 1000 × 0.000003, and returns 0.003. That's the level of code this guide uses — short, readable, traceable.

2. HTTP requests

Every LLM call in this guide is an HTTP request — the same mechanism your browser uses to load a page. Your code sends a request to a URL (an API endpoint), the provider's server does work, and sends back a response.

import requests

r = requests.post(                       # POST = "here's data, do something"
    "https://api.example.com/v1/chat",   # the endpoint
    json={"message": "Hello"},           # the request body
)
print(r.json())                          # the response, parsed

You'll see the words request, response, endpoint, and POST constantly. They mean exactly this.

3. JSON

JSON (JavaScript Object Notation) is the universal text format for structured data — every LLM API speaks it. It's just nested keys and values:

{
  "model": "some-model-name",
  "messages": [
    {"role": "user", "content": "Hello"}
  ],
  "max_tokens": 100
}

Curly braces {} hold key–value pairs; square brackets [] hold lists; strings are quoted. If you can read that block, you can read every API example in this guide.

4. API keys and environment variables

An API key is a secret password that tells the provider who's calling (and whom to bill). You never paste it into code — you put it in an environment variable (a named value your operating system holds for your programs):

export OPENAI_API_KEY="sk-..."   # set it in the terminal

import os
key = os.environ["OPENAI_API_KEY"]   # read it in code

Why it matters: keys committed to code get scraped by bots within minutes and your account gets drained. Environment variables keep secrets out of code.

5. The terminal

The terminal (command line) is the text interface where you run programs. The handful of commands this guide uses:

cd my-project        # move into a folder
python script.py     # run a Python file
npm install openai   # install a package (Node.js)
pip install openai   # install a package (Python)

Stage 0 of the roadmap walks through setting all of this up from a blank machine, step by step.

Why it matters

Every later page assumes these five ideas without re-explaining them. "Send the conversation as JSON over HTTP with your API key" should parse instantly by the time you reach Messages — if it does, nothing else in this guide will block on programming knowledge.

Common pitfalls

Where newcomers trip

• Skipping this check. If "parse the JSON response" sounds foreign, later pages will feel harder than they are. The fix costs a week now, not months later.
• Pasting API keys into code. The classic first-week mistake. Use environment variables from day one.
• Confusing Python and JavaScript examples. This guide shows both. You only need one — pick Python if unsure, and mentally skip the other.
• Thinking you need more than this. You don't need data structures, algorithms, or math to start. (When you eventually want the math, the optional math primer is the appendix for it.)

Never coded at all?

Take one beginner course first, then come back — this guide will read easily afterwards.

Go deeper (optional)

• Our companion modern web dev guide (sibling guide in this collection) owns programming-from-zero and teaches it the same way this guide teaches AI.
• Any "Intro to Python" course that gets you writing functions and running files from the terminal is enough.

🤔 Quick checkQuick check

---

→ Next: Tokens — the unit every LLM reads, writes, and bills by.