Version Control

Version control is a system that systematically stores and makes traceable the changes made to files. Each saved change is called a commit — a snapshot of the project at a specific point in time. These snapshots stack up into a seamless history. Anyone who introduces an error can undo it at any time by jumping back to an earlier state. This works just as well for a single person as it does for a team of hundreds of developers.

Why a project without version control is risky

Without version control, every change irretrievably replaces the previous state. Anyone who notices that their change caused a bug must search for it and undo it manually — without knowing where to start. In a team of ten developers working simultaneously on the same files, this quickly leads to chaos: changes overwrite each other, and no one knows anymore who changed what and when.

Version control solves this problem fundamentally. It stores not just the current state, but the entire sequence of all changes, including the name of the person who made them and a brief justification. This makes errors traceable, responsibilities clear, and projects maintainable in the long run.

Branches, commits, and merging changes

The core of modern version control is the concept of the branch. A branch is a parallel working copy of the project. Developer A works on a new feature in her branch, while developer B simultaneously fixes a bug in his own branch. Neither interferes with the other. Only once their work is finished and reviewed do the changes flow back into the shared main line. This step is called a merge.

During a merge, a conflict can occur: if A and B have changed the same line differently, the system must ask which version should take effect. This sounds like extra work — and in the short term, it is. In the long run, however, this procedure prevents changes from simply being silently lost. The by far most widely used version control system today is called Git. It was developed in 2005 by Linus Torvalds, the founder of the Linux operating system kernel.

Git works in a distributed manner: every developer has a complete copy of the entire project history on their own machine. This means one can create commits and browse the history even without an internet connection at the moment. A connection to the shared server is only needed when synchronizing with others.

Version control in products and AI projects

Platforms such as GitHub or GitLab use Git and build a user interface on top of it. They are simultaneously a storage location for billions of lines of code and a social network for developers. Anyone who wants to improve an open-source project — that is, a project with freely accessible source code — submits their proposed change as a so-called pull request. The project maintainers review it, discuss it publicly, and decide whether it gets incorporated.

Version control is also central in AI projects — though Git alone is often not enough there. A trained AI model consists of enormous files containing billions of numerical values, known as weights. These files are too large for classic Git workflows. That is why specialized extensions exist, such as Git LFS (Large File Storage), or dedicated platforms like Hugging Face, which manage model versions similarly to Git but are optimized for this type of file.

Version control often appears indirectly in news reports: when a company announces the release of a new version of its product, there is almost always a version control system behind it that has documented the entire development history of that version. Without it, coordinated software development at today’s scale would simply not be possible.