# Persistent Memory

QwenLM has released Qwen Code, an open-source AI coding agent whose entry points span the terminal, IDE plugins, a desktop app, a Web UI, SDKs and chat integrations including Telegram, DingTalk, WeChat, WeCom and Feishu. It ships with Auto-Memory, Skills, SubAgents and MCP, plus Agent Teams, Agent Arena and Dynamic Workflows, some marked experimental or opt-in. A daemon mode lets browser, editor, SDK and automation clients share one agent session, and clients can reconnect to active sessions.

QwenLM has released Qwen Code, an open-source AI coding agent whose entry points span the terminal, IDE plugins, a desktop app, a Web UI, SDKs and chat integrations including Telegram, DingTalk, WeChat, WeCom and Feishu. It ships with Auto-Memory, Skills, SubAgents and MCP, plus Agent Teams, Agent Arena and Dynamic Workflows, some marked experimental or opt-in. A daemon mode lets browser, editor, SDK and automation clients share one agent session, and clients can reconnect to active sessions.

Long-running LLM agents often write each episode's lesson straight into memory, yet local success is not a reliable sign of reusable, safe knowledge. FGSE turns every candidate lesson into a structured hypothesis with a precondition, behavior change, expected effect and verifier, applies it only in a temporary state, and tests it on target-transfer, falsification and archived-regression probes before committing, narrowing or rejecting it. On verifier-rich benchmarks including WebArena, AppWorld and τ-bench, FGSE with a DeepSeek-V3.1 backbone reaches 39.8 WebArena success rate and 78.8 AppWorld average completion, while cutting falsification failures, regression damage and harmful committed updates to 5.0%, 1.3% and 3.6%.

Long-running LLM agents often write each episode's lesson straight into memory, yet local success is not a reliable sign of reusable, safe knowledge. FGSE turns every candidate lesson into a structured hypothesis with a precondition, behavior change, expected effect and verifier, applies it only in a temporary state, and tests it on target-transfer, falsification and archived-regression probes before committing, narrowing or rejecting it. On verifier-rich benchmarks including WebArena, AppWorld and τ-bench, FGSE with a DeepSeek-V3.1 backbone reaches 39.8 WebArena success rate and 78.8 AppWorld average completion, while cutting falsification failures, regression damage and harmful committed updates to 5.0%, 1.3% and 3.6%.

MongoDB released and open-sourced an integration that turns MongoDB Atlas into the persistent memory store for AI applications and agents built with the Vercel AI SDK. The SDK itself ships no database, so an application previously started every conversation with no knowledge of the last one, and developers had to design schemas, write cleanup jobs and set up vector search themselves. The integration sorts memories into session, semantic, procedural, episodic and scratchpad types with separate retention policies, and creates vector indexes automatically while detecting embedding dimensions at startup.

MongoDB released and open-sourced an integration that turns MongoDB Atlas into the persistent memory store for AI applications and agents built with the Vercel AI SDK. The SDK itself ships no database, so an application previously started every conversation with no knowledge of the last one, and developers had to design schemas, write cleanup jobs and set up vector search themselves. The integration sorts memories into session, semantic, procedural, episodic and scratchpad types with separate retention policies, and creates vector indexes automatically while detecting embedding dimensions at startup.

claude-mem gives Claude Code persistent memory across sessions: it automatically captures tool-usage observations during a session, generates semantic summaries, and makes them available to future sessions, so Claude keeps continuity of knowledge about a project after sessions end or reconnect. Installation runs through npx claude-mem install or the plugin marketplace, and the installer configures everything before asking the user to sign in and choose a memory provider.

claude-mem gives Claude Code persistent memory across sessions: it automatically captures tool-usage observations during a session, generates semantic summaries, and makes them available to future sessions, so Claude keeps continuity of knowledge about a project after sessions end or reconnect. Installation runs through npx claude-mem install or the plugin marketplace, and the installer configures everything before asking the user to sign in and choose a memory provider.

Memento sends sensor observations by reference: the device searches a vector table in persistent memory for a similar past sample and, on a hit, transmits only an index, and modeling shows that on low-rate links such as LoRa a 25% hit rate cuts latency and energy by roughly a quarter.

Memento sends sensor observations by reference: the device searches a vector table in persistent memory for a similar past sample and, on a hit, transmits only an index, and modeling shows that on low-rate links such as LoRa a 25% hit rate cuts latency and energy by roughly a quarter.

The work introduces LiveMACEBench, a process-aware benchmark that uses live cryptocurrency and U.S. equity markets as a naturally evolving testbed in which five frontier LLMs run continuous paper-trading trajectories under matched Base, Tool Use, Persistent Memory, Rule Following, and Multi-Agent Collaboration configurations, evaluated through both realized outcomes and trace-derived diagnostics; across roughly 30 days it finds a pronounced outcome–capability gap, with return rankings fluctuating and often diverging from capability-specific measurements while similar outcomes arise from markedly different patterns of mechanism use.

The work introduces LiveMACEBench, a process-aware benchmark that uses live cryptocurrency and U.S. equity markets as a naturally evolving testbed in which five frontier LLMs run continuous paper-trading trajectories under matched Base, Tool Use, Persistent Memory, Rule Following, and Multi-Agent Collaboration configurations, evaluated through both realized outcomes and trace-derived diagnostics; across roughly 30 days it finds a pronounced outcome–capability gap, with return rankings fluctuating and often diverging from capability-specific measurements while similar outcomes arise from markedly different patterns of mechanism use.