Ideas don't stay small. A single concept branches into sub-concepts. A question leads to a dozen related questions. Your thinking expands, and now your maps can expand with it.
Submaps
Every node in your map can now become its own map.
Double-click any node and you're inside it, looking at a new canvas where that concept's sub-ideas live. The main map stays clean. You see "Machine Learning" as a single node, while the details (supervised vs. unsupervised, neural networks, training data) live one level deeper.
Think of it like folders, but for thinking. Your high-level map shows the big picture. Each node contains a world of related ideas you can zoom into when you need depth and zoom out when you need perspective.
Breadcrumbs at the top show where you are in the hierarchy. Navigate back up anytime. The structure emerges naturally as your understanding grows.
Autocomplete
Start typing and Atlas suggests how to complete your thought.
The suggestions aren't generic. They're trained on your existing notes and maps: your vocabulary, your phrasing, your patterns of thinking. When you're halfway through a sentence, the completion feels like something you would have written.
Press Tab to accept a suggestion. Keep typing to ignore it. The feature stays out of your way until you want it, and it's surprisingly good at predicting what comes next.
This isn't about writing faster (though it is faster). It's about reducing the friction between thinking and capturing. When the right words appear before you finish typing them, you stay in flow.
10 New Map Layouts
Not every idea fits the same shape.
Some concepts are hierarchical. They flow from general to specific, parent to child. Others are relational: a web of connections where no node is more important than another. Some radiate from a center; others cluster in groups.
We added 10 layout algorithms to match how ideas actually organize:
- Hierarchy. Classic tree structure, like an org chart. Top to bottom, general to specific.
- Relationship. Emphasizes connections between nodes. Good for showing how concepts influence each other.
- Radial. One central idea with related concepts radiating outward. Perfect for exploring a core topic.
- Plus seven more: dagre, elk, force-directed, circular, breadth-first, grid, and concentric.
Switch layouts anytime. Your content stays the same, only the arrangement changes. Try different views until you find the one that reveals the pattern you're looking for.
Improved
Several features got meaningful upgrades:
-
Speech-to-text. Click the microphone icon in chat and dictate your thoughts. Useful when you're thinking faster than you can type, or when your hands are full.
-
Embeddings view. See which sources informed each part of your map. Understand the provenance of ideas: where they came from, what evidence supports them.
-
Contextual zoom. Maps now zoom smoothly to focus on what you're working on. Select a cluster of nodes and the view adjusts to frame them perfectly.
-
Layout stability. Adding new nodes no longer causes the whole map to jitter and rearrange. Nodes stay put while you build around them.
-
Citation links. Open correctly in all browsers. Click a citation and go straight to the source.
Have feedback? We'd love to hear from you at team@atlasworkspace.ai