Memory & Self-Improvement
Limaze AI lims learn and improve over time — even when you're not using them.
Lim Memory
Every lim maintains persistent memory across all sessions:
- Conversation history — what was discussed and decided
- Task outcomes — what worked and what didn't
- Learned preferences — your style, feedback patterns, and corrections
- Knowledge notes — facts and insights gathered over time
Memory persists across sessions. A lim that wrote 50 blog posts for you will be noticeably better at post #51 than it was at post #1.
DreamTask (Self-Improvement Loop)
Every 6 hours, lims enter a "dream" cycle where they:
- Review their recent work and feedback
- Identify patterns and areas for improvement
- Generate new skills from experience
- Update their operating strategies
Lessons & Proposals
The learning system generates two types of improvements:
Lessons
Patterns learned from experience — e.g., "Tweets with questions get 40% more engagement for this brand." Lessons can apply at the office, lim, or capability level.
Improvement Proposals
Concrete suggestions for upgrades:
- New Capability — "I should learn how to create Instagram carousels"
- Upgrade Capability — "My headline writing could use A/B testing"
- Refine Lim Package — "Adjust my tone to be more conversational"
Proposals go through a pipeline: queued → approved → generating → testing → applied. You can review and approve them before they take effect.
Memory Policy
You can customize how each lim handles memory:
- What types of information to remember
- How long to retain data
- What to prioritize vs. forget
- Privacy boundaries (what not to store)