How OpenNote Scaled Personalized Visual Learning with Mem0 While Reducing Token Costs by 40%
Mem0 turned our AI tutors into true learning companions - tracking each student’s struggles, strengths, and learning style across the entire platform and tools.
- Abhi Arya, Co-Founder, Opennote
About OpenNote
OpenNote is an AI-powered learning platform built for students who prefer a visual and interactive approach to studying. The platform combines diagrams, explanations, and conversation into a unified workspace that supports asynchronous learning. With built-in AI support, OpenNote delivers personalized guidance that helps users navigate difficult topics, organize resources, and build lasting understanding.
The Challenge: Bringing Continuity to Open-Ended Learning Flow
OpenNote was designed for exploratory learning. A student might highlight part of a diagram, revisit the topic days afterward, and ask questions at any point in between. But this flexibility exposed a key limitation: the AI had no memory of what the user had seen, asked, or saved.
The team identified three critical challenges:
- Disconnected sessions: Every time a student returned, the AI could not recall previously asked questions or learning paths.
- Heavy prompts: Including full interaction history in each prompt increased token usage and response latency.
- Product complexity: Building a custom memory infrastructure would have delayed roadmap milestones and diverted engineering focus.
OpenNote needed a memory solution that could support adaptive learning across sessions without requiring heavy development work.
The Solution: Integrating Mem0 for Session-Aware AI
OpenNote chose Mem0 to bring persistent, session-aware memory to its AI layer. It made it possible to create learner profiles that evolved over time and retained relevant context across conversations.. Mem0 was integrated within days, with no need to modify core infrastructure.
“We’re incredibly glad to be working with this team and have Mem0 built into Feynman-2, our AI tutoring engine, driving learner profiles on OpenNote. ”
- Rishi Srihari, Co-Founder, Opennote
With Mem0, OpenNote was able to:
- Retain prior student interactions to improve follow-up accuracy
- Generate smarter prompts with only the most relevant summaries, reducing token usage
- No custom logic required to summarize or retrieve user history
The AI assistants could now respond with continuity and relevance, even across long, fragmented study sessions.
Results: Higher Engagement, Lower Cost, Less Engineering Overhead
OpenNote observed clear improvements in both user experience and system performance after integrating Mem0.
| Metric | Before Mem0 | After Mem0 | Change |
|---|---|---|---|
| Token Usage per Prompt | Full history included | Summarized, relevant memory | 40% reduction |
| Session Continuity | Stateless, repetitive chats | Personalized and coherent conversations | Increased engagement |
| Engineering Effort | Estimated 3-4 weeks build | Integrated in 2 days | Fast Implementation |
| User Feedback | Inconsistent experience | Feels personalized | Positive qualitative shift |
Personalized Learning in Practice
With memory in place, Feynman-2 and OpenNote’s other AI assistants became a more capable learning companion. Students received support that reflected their prior progress and questions.
Examples included:
- After a session on probability:
“You highlighted the chain rule diagram yesterday. Want to apply it to this example?” - Returning to a physics lesson:
“You stopped at Newton’s Second Law. Would you like a quick recap before moving to momentum?” - Reviewing a concept previously marked as difficult:
“This video connects to a topic you struggled with last week. Would a simplified version help?”
These responses helped students stay engaged, avoid repetition, and build knowledge progressively.
“With Mem0, we’re able to work towards our goal of providing the best personalized experience for students”
- Vedant Vyas, Co-Founder, Opennote
These subtle improvements created a more natural, coherent experience that helped students stay engaged, avoid repetition, and build knowledge progressively.
Why It Matters
For OpenNote, learning is not linear. Students move across topics, formats, and questions at their own pace. Mem0 made it possible to support this flexibility with a memory system that scales and adapts. By integrating Mem0, OpenNote improved tutoring quality, reduced infrastructure cost, and delivered a better experience to students without slowing down product development.
The result is a smoother, smarter learning journey that helps students stay engaged and make meaningful progress - one session at a time.
Interested in giving your AI a real memory?
Explore Mem0 at https://mem0.ai