Forge

In the end, a program consists of written text, the so-called source code. This text is often written by many people at the same time, in different places around the world. So that this doesn’t end in chaos, they need a shared storage location with clear rules. That is exactly what a forge is: a website where the source code resides and every change is recorded traceably. Added to this are tools for collaboration, such as lists of open bugs and discussions about planned changes. The English word “forge” refers to a smithy — here, software is forged.

Why software is created in a central place today

20 years ago, developers often sent their changes around by email. That worked for small projects, but not for thousands of participants. Forges have standardized this work. Anyone who wants to contribute to a large project today finds the same workflows everywhere.

This has turned forges into a part of the infrastructure on which the entire economy depends. Almost every app and every website uses third-party building blocks that reside on a forge. If a major forge goes down, development stalls worldwide. And whoever operates a forge can see what the industry is currently working on.

That’s why forges are also economically interesting. Microsoft bought GitHub in 2018 for 7.5 billion dollars. That wasn’t the purchase of a website, but access to a community of around 28 million developers at the time. Today there are more than 100 million registered accounts.

Version control, branches, and pull requests

The heart of a forge is version control. It stores not only the current state of the code, but every single change with date and author. You can jump back at any time and see what a file looked like last year. The underlying program used is almost always Git. It’s important to distinguish: Git is the tool on your own computer, the forge is the server with the web interface around it.

So that two people don’t get in each other’s way, everyone works in their own branch. A branch is a copy of the project in which you can freely experiment. Once the change is finished, you submit a request to merge it. On GitHub this is called a pull request. Other participants read the proposal, comment on individual lines, and finally approve it.

Additionally, automatic checks run. As soon as someone submits code, the server starts tests and checks whether the program still works. If it finds an error, the proposal is flagged and not merged. This automation is called continuous integration.

GitHub, GitLab, and the role of AI

The best-known forge is GitHub, followed by GitLab. Those who value European servers or free software often use Codeberg or a self-hosted GitLab instance. Many government agencies and corporations also install forges on their own machines so that internal code doesn’t leave the building.

In the news, forges mainly come up in connection with AI. The huge code collections there were the training material for programming assistants like GitHub Copilot. Whether that was permitted is legally disputed, because many projects are under licenses that impose conditions on reuse. At the same time, GitHub itself is building AI features directly into the platform.

Incidentally, for a student, a forge is not an expert tool. A free account is enough to store your own projects or report a typo in someone else’s project. Many companies look at exactly this public profile when reviewing job applications.