Ever lost hours of work to an accidental delete? Closed a tab and watched your draft vanish? We've been there too. This week, Atlas learns to remember everything so you never have to worry about losing your work again.
Version History
Your notes now keep a complete history of every change. Click the history icon in the toolbar to open a timeline of every saved version, each with a timestamp showing exactly when you made it.
But we didn't stop at simple backups. Before you restore anything, you can preview the old version side-by-side with your current draft. Word-level diffs highlight exactly what changed (additions in green, deletions in red) so you know precisely what you're getting back.
Worried about overwriting something important? Don't be. When you restore an old version, Atlas automatically saves your current state as a new version first. It's restoration without risk. You can always go back to where you were.
Behind the scenes, versions are stored using diff-based compression. That means your workspace stays fast no matter how many drafts you've written or how many times you've revised that thesis chapter.
Message Branching
Conversations with Atlas aren't linear anymore. Sometimes you ask a question one way, get an answer, and wonder: what if I'd asked it differently?
Now you can find out. Edit any message you've already sent and Atlas creates a new branch, a parallel thread that explores a different direction while preserving your original conversation intact.
Think of it like version control for thinking. You can compare how different questions lead to different answers, keep the branches that work, and let the others fade into the background. Each branch maintains full context, so the AI knows everything that came before.
This transforms chat from a one-shot Q&A into a genuine thinking tool. Explore alternatives. Test hypotheses. Follow tangents without losing your main thread.
Reasoning Traces
When Atlas cites a source in its answers, you can now see exactly why. Hover over any citation badge and a reasoning trace appears, showing which part of the source was relevant and how it connects to your question.
This matters because citations shouldn't be black boxes. If Atlas says "According to your research paper on page 12..." you should be able to understand the logic: why that page, why that passage, why it's relevant here.
Reasoning traces make verification faster. They help you trust (or question) the AI's answers. And they make Atlas more useful as a research partner, not just an answer machine, but a transparent collaborator that shows its work.
Improved
This release also includes refinements across Atlas:
-
Full-screen maps. Click the expand button on any map to enter full-screen mode. New zoom controls let you focus on specific clusters or pull back for the big picture. Maps are more useful when you can actually see them.
-
Formatting toolbar. The note editor gained new tools: task lists with checkboxes for tracking todos, dashed lists for informal outlines, and horizontal rules to visually separate sections. Select text to see which formatting is active and toggle styles without touching markdown.
-
Performance. We cached workspace data more aggressively. Opening your projects, switching between notes, loading the sidebar: everything feels snappier. The landing page no longer flickers when you're logged in; you go straight to your last project.
-
Links as badges. When you reference another note in your writing, the link now appears as a clean badge. It's easier to see connections at a glance.
-
Version history panel. Now updates immediately after you restore a version, so you always see the current state.
-
Formatting toolbar sync. Selection state stays accurate as you type and move around your document.
-
PDF processing. Maps regenerate reliably when your source documents change.
Have feedback? We'd love to hear from you at team@atlasworkspace.ai