Mandol: A Memory System for Long-Term Chats
jiqizhixin · x · 2026-07-13
Researchers introduced Mandol, a memory system designed for long-range conversations, aiming to solve the issue of AI "forgetting" after a few weeks.
Core Concept
- Unifies fragmented data into a semantic graph
- Avoids the inefficiency of cross-database I/O and retrieval noise
- Makes long-term memory more suitable for agentic systems
Results
- Outperforms existing agents on long-term conversation benchmarks like LoCoMo and LongMemEval
- Achieves a 5.4x speedup in retrieval speed on consumer hardware
- Achieves a 4.8x speedup in insertion speed
- Tested under 10 QPS conditions
The post also includes links to the paper, project page, and mechanism report.
More from coding & agent
- AI agents are starting to strain code hosting platforms — craigsdennis · 2026-07-21
- Async OPD distillation doubles throughput while matching synchronous math accuracy — _lewtun · 2026-07-21
- Omnigent 0.6.0 adds Claude Code imports, Slack approvals and desktop apps — matei_zaharia · 2026-07-21
- Google appears to have quietly shipped Gemini 3.6 Flash, with lower pricing and better agentic scores — xiaohu · 2026-07-21
- Open-source CLI audits AI tools, MCP configs, and agent skills on local machines — Initial-Copy332 · 2026-07-21
- Coding agents feel less stressful when the 5-hour limits are temporarily removed — iamrobotbear · 2026-07-21