Code
A chatbot can write code, even for people who can't program. Like a fast handyman who builds you a shelf. Before you put your books on it, you test that it holds. Learn how to make something small with AI, and when to ask an expert.
First, what do you think?
A chatbot writes a small program for you, and it runs. Is it safe to use? Have a guess.
What can AI do with code?
The whole story in three sentences. Code is text, so a chatbot can predict it. From your description, it writes a small program, a formula or a web page. That helps beginners and experts, but you still test what it builds.
Think of a fast handyman. You describe the shelf you want. Ten minutes later, it's on the wall. Great. But before you put your books on it, you push on it. Does it hold?
- Write. A small game, a web page, a formula that adds up your expenses.
- Explain. "What does this line do?" A chatbot can explain code step by step.
- Fix. Paste an error message, and ask what went wrong.
How much does it help? Research gives a mixed picture.
- Faster on a fresh task. In one study, developers had to build a small web server. The group with an AI assistant finished 55.8% faster.
- Slower on familiar work. In another study, experienced developers worked on their own large projects. With AI they took 19% longer. They thought it had made them faster.
- Less secure. In a third study, people with an AI assistant wrote less secure code. And they were more likely to believe it was secure.
What the predictor does
The predictor writes code that fits your description. That's the rule.
Code that fits isn't always code that's right. It may work for your first test, and fail for the next. It may work, but leave a door open for hackers. The predictor doesn't test its own work unless a tool makes it.
And there's a trap. Code that runs feels finished. That's why people in the third study felt safe while their code wasn't. The ones who trusted the AI less did better.
The handyman builds fast. You push on the shelf before you trust it. For anything with passwords, money or personal data, you ask an expert to push too.
Try it yourself
Three moments when a chatbot wrote code for you. What's your next step?
Everything runs in your browser. Nothing you type is sent anywhere.
How do you build something with AI?
Four steps. Try them today on something tiny, like a page with a joke of the day.
Describe it in plain words
Say what it should do, for whom, and where it runs. Start small: one piece at a time.
Run it and test it
Try normal input, and odd input. An empty box, a very long name, a zero. Does it still work?
Ask it to explain
Ask: "Explain what each part does." Read along. You'll learn, and you'll spot parts that seem wrong.
Get a check for anything risky
Passwords, payments, personal data: let someone who knows security look at it. Never paste real passwords or keys into a chatbot.
In practice
Three people get help from the handyman.
Case 1The quiz game
Ruben (15) wants a quiz game about football for his friends. He's never programmed. A chatbot writes it in a few minutes, and it works.
Share it, and move on.
It's fine to share. But Ruben learns nothing about how it works. And for a school assignment, AI may help, but not replace his own work.
Ask the chatbot to explain each part. Change something small himself, like the colours or the questions.
The game works, and Ruben understands a bit of code.
Case 2The spreadsheet formula
Linda (38) needs a formula that adds up hours per project. A chatbot gives one. It looks right.
Use it for the whole team.
It may work for the first rows and go wrong for others. Code that fits isn't always code that's right.
Test it on a few rows she can check by hand. Including an empty cell and a weird date.
She finds the problem before the team relies on it.
Case 3The web shop
Erik (45) sells handmade candles. He asks a chatbot to build a shop page with payments. It runs on his laptop.
Put it online.
People with AI help write less secure code, and still feel safe. A payment page needs to be secure.
Use an existing shop service for payments. Or let an expert check the code before it goes online.
His customers' money and data stay safe.
When not to
- Not for passwords, payments or personal data, uncheckedIt can leave doors open.Use a trusted service, or ask an expert.
- Not instead of learningAt school, AI helps but doesn't replace your work.Ask it to explain, and write parts yourself.
- Not when you can't test itYou can't see if it's right.Start smaller, or ask someone to test it with you.
By situation
Use AI to explain code and errors. Write the parts that are graded yourself, and say when you used AI. Ask your teacher what's allowed.
Formulas, small scripts and quick tools save time. Test them on data you can check. Anything with customer data or money: ask IT first.
A page for your club, a small game, a tool for your hobby. Build it in small steps, test it, and keep real passwords out of the chat.
Myths and mistakes
| Claim | True? | Why |
|---|---|---|
| AI always makes programming faster | No | One study: 55.8% faster. Another: experienced developers 19% slower. |
| If it runs, it's safe | No | People with AI help wrote less secure code, and felt safer. |
| You need to know programming to start | No | You can start, but you do need to test. |
| You'll notice if it slows you down | Not always | Developers who were slower thought they'd been faster. |
Related topics
Read first
- Text and chatcode is text, so the same steps help.
Next
- AI agentsAI that doesn't only write, but also does things.
Related
- Privacy and your datawhat not to paste into a chatbot.
Sources
- researchPeng e.a. (2023). The Impact of AI on Developer Productivity: Evidence from GitHub Copilot.Developers had to build a small web server in JavaScript. The group with an AI coding assistant finished 55.8% faster.
- researchBecker e.a. (2025). Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity.16 experienced developers worked on their own large projects. With AI tools they took 19% longer. They expected to be 24% faster, and afterwards thought they had been 20% faster.
- researchPerry e.a. (2023). Do Users Write More Insecure Code with AI Assistants?.People with an AI assistant wrote significantly less secure code, and were more likely to believe their code was secure. Those who trusted the AI less did better.
- organisationKennisnet. Schoolafspraken over het gebruik van generatieve AI.Principles for schools: people come first, use only tools the school approved, be open about using AI, AI helps but never replaces your own work, protect privacy. Under 13 only with an adult.
Test
You get 7 questions about the whole page. Get 5 right, and you've mastered the topic.
About the tools in this course. Every tool is described in the same way, in alphabetical order, without a recommendation. Names, prices and features change often: the maker's own site is always right. This site was built with the help of Claude, one of the tools described here.