Repo
A repo is the central storage location of a software project: it contains all files plus the complete history of every change. This allows many people to work on the same software simultaneously without destroying each other's work.
Software ultimately consists of text files. These files don’t simply sit in a folder on a laptop, but in a repo. The short form stands for repository, meaning a place of storage. A repo contains all the files of a project plus a complete log of who changed which line and when. You can think of it as a folder with a built-in memory. Every old version is preserved and can be restored at any time.
Why software could no longer be created without repos
Larger programs are rarely worked on by individuals alone. For an operating system or an AI framework, hundreds or thousands of people are involved, often spread across many countries. Without a shared storage location, they would constantly overwrite each other. The repo is the one binding truth about what the current state of the project looks like.
Just as important is the reverse direction. If an update breaks a program, the log can be checked to see which change was responsible. It can be reverted individually without losing everything else. This ability lowers the risk of trying out something new in the first place.
For the public, repos have another significance too. With open-source software, whose code anyone may inspect, the repo is the place where you can actually verify what a program does. A company’s claims about its software can be checked there against the actual code.
Commits, branches, and the way back
The software that manages repos is, in almost all cases, called Git. Whoever has made a change bundles it into a commit, meaning a saved intermediate state with a short description. These commits line up to form a chain. It is the history of the project and can be traced back step by step.
So that nobody gets in each other’s way, people work in branches. A branch is an offshoot: a separate copy of the project in which you can experiment freely. Once the new feature is finished and tested, the branch is merged with the main state. This step is called a merge.
A repo usually exists twice over. On a server sits the shared version, and on every participating machine a complete local copy including the history. You fetch other people’s new commits with pull and send your own back with push. A pull request is the polite version of this: you propose a change, others review it and agree before it lands in the main state.
GitHub, Hugging Face, and leaked code in the news
The best-known platform for repos is GitHub, which has belonged to Microsoft since 2018. Millions of projects reside there, from school assignments to Linux. For AI models, Hugging Face has established itself as a similar go-to place. There, a repo contains not just code but also the trained model files themselves.
In business news, repos mainly come up on two occasions. First, at releases: when a company discloses a model, the repo reveals which license applies and exactly what is being shared. Second, in security incidents, when internal code from a private repo leaks to the outside.
A common misconception: a repo is not a cloud storage service like Dropbox or Google Drive. There, only the current file matters; here, the entire history matters. And there’s another possible mix-up in the finance section of the same news site. In banking, repo refers to a repurchase agreement involving securities. This term has nothing to do with software.