Git Atlas is a knowledge project in progress. It is less concerned with “how do I spell the command” and more with “what does this command actually change in Git's object model?”
Most Git tutorials organize themselves around commands. Readers quickly memorize reset, rebase, and cherry-pick, yet still do not know which step they moved on the timeline. Git Atlas wants to put those operations back on the same map.
A map with gravity
The project breaks Git knowledge into several kinds of nodes:
- objects: commit, tree, blob, tag;
- references: HEAD, branches, remote-tracking branches;
- actions: commit, merge, rebase, reset, stash;
- problems: recovering a lost commit, cleaning history, restoring a worktree.
Edges describe causality. reset --soft, for example, changes a reference but leaves the index and working tree alone. rebase creates new commit objects, which is why the old commits can still be found through reflog.
Why Markdown
A content-first project does not need a complex frontend on day one. Git Atlas starts as a knowledge base of Markdown and small frontmatter blocks. Scripts generate the node index and relationship graph.
type: command
command: reset
affects: [HEAD, index, worktree]
related: [reflog, restore]
Each piece of knowledge can be reviewed, linked, and reused on its own. The visual layer can change. The object model and the explanations underneath must stay stable.
Current state
The project is still building its data model and visual prototype. The next step is a command relationship map, followed by an interactive timeline demonstration.
It does not aim to replace the official Git documentation. It is a star chart for getting oriented when you are lost — a reminder of which direction to look.