Most version-control systems store a list of changes (diffs) per file over time. Git is different: each commit is a full snapshot of every file. To stay efficient, files that did not change between snapshots are not re-stored — both snapshots simply point to the same content. That single idea — content stored once and referenced by many — is the heart of Git, and it is enabled by content addressing.