Claude Code and Codex are both suitable for your first application, and for beginners what they have in common matters more than how they differ. Base your decision on three practical questions: do you already have an account with one of the providers, does the tool run on your computer, and which way of working suits you?
This post shows you how to answer those questions and why the Vibe Coding Foundation course allows both agents. You will deliberately not find a feature comparison here. Both tools are under constant development, and what a comparison table says today may look different in a few months.
What Claude Code and Codex have in common for beginners
Both are AI agents. You describe a task in plain language, the agent suggests code or changes files in your project, and you check the result. That is exactly what vibe coding means: developing software with the help of AI.
If you have no programming experience, you will run into the same hurdles with either tool:
- Vague instructions lead to vague results. "Make the app better" gets you something or other. "Show only contacts with the status open in the contact list" gets you something you can check.
- The agent can be wrong. It writes code that seems to work at first glance but fails to catch an empty input or saves a status change incorrectly.
- English comes with the territory. Code and many tool messages are in English, whichever agent you use.
- You remain responsible. Whatever ends up in your project, you have to be able to explain it in the end.
Whether your first app works out depends mainly on how you phrase tasks and check results. The tool is the smaller part of the equation.
What matters when you choose
Do you already have an account with a provider?
This is often the easiest decision. An account you already use means one sign-up less, one bill less and an environment you know. Which account types include access to each agent, and on what terms, is up to the provider. These terms change, so it is best to check them directly with the provider before you start.
If you are learning for your job, ask internally first. Some companies already use one of the two agents or have rules on which AI tools are allowed. In that case the choice is made quickly.
Does the tool run on your computer?
For the course you need a laptop or desktop running Windows, macOS or Linux. A tablet is not enough. You also need Node.js version 20 or later, Git and permission to install software. That last point is easily overlooked on company laptops.
Whether and how Claude Code or Codex runs on your operating system is described by each provider in its own instructions. Read them before you commit. If installing one tool on your device seems awkward, that is a good reason to pick the other.
Which way of working suits you?
There is no right or wrong here, only different preferences. Some people prefer to give small instructions and check each step individually. Others like to describe a larger task and then look at what has been produced. For beginners, small steps are usually calmer, with either tool.
If you have access to both, give them the same small task, for example: "Add a field for the next follow-up date to every contact." Then pay attention to three things: Do you understand the answer? Can you see which files were changed? Do you find it easy to ask a follow-up question when something is unclear? The tool for which you can more readily answer yes is the better fit for you.
Checklist before you decide
- I know whether I already have an account with one of the providers.
- I have read the provider's current terms for access.
- My operating system is supported according to the provider's instructions.
- I am allowed to install software on my computer.
- Node.js version 20 or later and Git are installed, or I know how to install them.
- If I am learning for my company: I have clarified which tool is allowed there.
What matters less than it seems
Online you will find rankings, comparison tables and forum posts that put one tool or the other clearly ahead. They do little to help your decision. Many of these comparisons are about experienced developers with large projects, and they go out of date quickly.
A contact and follow-up app with a list, a detail view, a filter and status changes is a clearly defined task. For that you do not need a tool that won a specialist benchmark. You need one that runs reliably on your machine and that you enjoy working with.
Why the course allows both agents
The skills the programme certifies are not tied to any particular tool. You learn to break a work task down into verifiable steps and describe them as acceptance cases. You build a simple application with an AI agent and change it in a traceable way. You test failure cases, data validation, access and secrets. And you explain your solution, name its limits and apply it to a new requirement.
What is assessed is the result, not how you got there. Every mission has criteria in the form of tests. On your computer a command checks these criteria, but what counts is the check on the academy's server, in an isolated environment with no network access. The tests look at whether your app rejects an invalid input or responds sensibly to an empty response from the simulated interface. Which agent wrote the code makes no difference.
So that it stays clear what comes from you, you keep an AI log: what did the AI contribute, and what did you change yourself? The written project exam also asks about your project and your decisions, not about the quirks of a tool. How it works is explained under exam procedure and rules (in German), and you will find the points breakdown in the exam rubric.
How Vibe Coding Foundation supports your start
The course begins with Mission 0, which takes around an hour. There you prepare your computer, your AI access and the starter project, and launch the app for the first time. The mission is freely available: under Mission 0: setting up your tools you can see what to expect before you buy. After that you build out the app step by step, from the contact list through the simulated interface to exporting overdue follow-ups as a CSV file. All missions with planned times are listed in the course overview.
In total, plan for around 30 hours, for example six weeks at about five hours a week. You learn online and at your own pace, with no fixed dates. If you get stuck, hints in three levels help you move on.
The course and project exam cost CHF 199 one-off including VAT, with three months of access and one exam attempt. You bring your own access to Claude Code or Codex, and its cost comes on top of the course price. Details of what is included are under price and services.
Can you switch tools later?
Your project is a folder of files whose history is tracked by Git. Switching agents therefore does not mean starting over. The new tool carries on working on the same files, and the tests still check the same behaviour.
Even so, for beginners it is calmer to stick with one tool during the course. Every switch brings a new installation, new messages and a new adjustment period, and that distracts from the actual learning. If you do switch, record it in your AI log. That way you can explain later which part of your app was built with which agent.