Commit

Software consists of many text files that are constantly being changed. So that no one loses track, developers record every step of their work individually. Such a recorded state is called a commit. A commit stores which lines were added, changed, or deleted in which files. It always includes a short description, a timestamp, and the name of the person who made the change. You can think of it like a save point in a video game, except you can jump back to any earlier point at any time.

Why projects without save points end in chaos

Large programs are often worked on by hundreds of people at the same time. Without commits, no one would know who wrote which line. If a bug then turns up, there would be no way to find its cause. With a chain of commits, on the other hand, it can be determined exactly from which point onward a program no longer works correctly.

Commits also make it possible to undo mistakes. If a website breaks after a change, the team simply jumps back to the last working commit. That takes seconds instead of hours. This is exactly why teams dare to try risky changes in the first place.

For companies, commits also serve as proof. They show when a security vulnerability was closed and who was responsible for it. In regulated fields such as medical technology or banking, such a complete record is even required by law.

What actually gets saved when you commit

The work is handled by a version control system, i.e. a program that logs changes to files. By far the most widespread one is called Git. First you select which changes belong together. Then you write a commit message, for example: “Login form now validates the email address.” Only then is the state finally recorded.

Every commit is given a unique identifier made up of letters and numbers, such as a3f9c2b. This identifier is calculated from the content of the change. If someone alters even a single character afterward, the identifier no longer matches. This means that retroactively falsifying the history is immediately noticeable.

Importantly: a commit does not save the entire project anew, but only the difference from the state before it. Every commit references its predecessor. This creates a long chain that can be traversed in both directions. A common misconception is that a commit is already published. In fact, it initially exists only on one’s own computer and has to be uploaded separately.

Commits in the news, on GitHub, and in AI tools

Commits are most visible on platforms like GitHub or GitLab. There, anyone can view the change history of open projects. For well-known programs like the Linux kernel, this now amounts to over a million commits. Developer profiles on these platforms also show how many commits they have contributed per day.

In tech news, commits often serve as an early warning system. Journalists discover hints of yet-unannounced products or features in public commits. Conversely, after a hacker attack, it can be checked retroactively when suspicious code was introduced.

For AI, the topic is doubly relevant. Models capable of writing code were trained to a large extent on publicly viewable commits. And modern coding assistants now suggest commit messages themselves or generate entire commits. Anyone who adopts such suggestions remains responsible nonetheless: their own name ends up under the change.