One of the questions we get most often is: “How does Ditto actually remember?” This post explains the memory systems behind Ditto.
Traditional AI vs. Persistent Memory
The Limitation of Stateless AI
Most AI assistants run in what we call “stateless” mode. Each conversation stands alone:
- No context from previous conversations
- No record of your preferences
- No knowledge built up over time
So you re-explain your context, preferences and goals at the start of each chat.
Ditto’s Memory Architecture
Ditto uses a memory system with several layers:
Short-term Working Memory
- The current conversation
- The task in front of you
- Preferences as you state them
Long-term Episodic Memory
- Past conversations and recurring patterns
- Project timelines and milestones
- Personal preferences and how you work
Semantic Knowledge Graph
- What you’ve taught it about your field
- How concepts relate to each other
- People, projects and topics in your work, updated as they change
Privacy and Security
Persistent memory raises fair privacy questions. Here is what you control:
- No model training: Your memory is not used to train AI models
- Temporary chats: Start a chat that is not saved to memory
- User Control: You can see, edit, export or delete any memory at any time
- Selective Recall: Choose what each message draws on from your memory
The Future of AI Memory
Next on our roadmap:
- Collaborative Memory: Shared knowledge across team members
- Cross-Platform Sync: The same memory on all your devices
- Proactive Insights: Suggestions based on patterns in your memory
Experience the Difference
Want an AI that remembers your context? Try Ditto today and see what persistent memory does for your work.
Want the technical details? Read our developer documentation on how the memory system works.