GitHub

Software consists at its core of text files. A human writes instructions down, and the computer later executes them. GitHub is an internet platform where such text files are stored and jointly edited. Each project resides there in its own folder, called a repository. What’s visible is not just the current state, but the complete history of changes: who changed which line when, and why. Millions of programmers work together this way, even though they never meet. GitHub was founded in 2008 and bought by Microsoft in 2018 for around 7.5 billion dollars.

GitHub as the address of the open-source world

Many programs are open source. That means: the program text is public, anyone may read, copy, and improve it. Almost the entire open-source world has gathered on GitHub. Anyone looking for a tool for their own work therefore looks there first.

This applies especially to artificial intelligence. When a company publishes an AI model, the associated program code usually appears as a GitHub project. The number of stars, meaning the public bookmarks of other users, is considered a rough measure of popularity. Projects with a hundred thousand stars regularly appear in tech news as evidence of a trend.

For individual developers, GitHub has a second function. The profile publicly shows what someone has contributed to. HR departments in the software industry often look at this, similar to a portfolio in a design job application.

Version control, branches, and pull requests

Underneath GitHub is a program called Git. Git stores not just the latest version of a file, but every intermediate state. Such saved intermediate states are called commits. You can go back to an older commit at any time if a change has caused damage. Git runs on your own computer; GitHub is the shared storage location on the internet.

So that several 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 one can experiment freely. Once the work is done, one submits a pull request: a request to merge the changes into the main version. Other contributors read the change, comment on it, and eventually approve it.

In addition, there are issues, meaning collected bug reports and requests. Any user can report that something isn’t working. Automatic checks can also be set up. With every pull request, a server tests whether the program still starts and no known errors come back.

Copilot and its role in the AI debate

GitHub is also important for people who never program themselves. A large portion of the AI tools that get reported on are published there. When a news report says a model has been released as open source, the proof can be found in a repository.

Also well known is GitHub Copilot. This is an AI assistant that makes suggestions while writing programs, similar to the word suggestions on a phone. Copilot was trained on huge amounts of public program code, among other sources from GitHub itself. This is exactly where a dispute hangs: some developers consider it impermissible that their code was used as training material without permission. There have been lawsuits about this in the USA.

There’s a common mix-up worth knowing about. Git is the tool for version control, GitHub is just a provider that makes it usable on the internet. Competitors like GitLab or Codeberg work on the same principle. Anyone who wants to can run Git entirely without a platform on their own server.