Tony Zhang, former DeepSeek V1 author and DeepSeek-VL project lead, unveiled Linger, an open-source personal agent that runs on the desktop rather than as an app. Its core is a local memory hub called Another Me that nightly distills email, calendar and chat into a personal wiki, scoring 95% on the LoCoMo long-term memory benchmark, while supporting third-party and locally optimized LLMs.
Long-term Memory
长期记忆
A mechanism enabling LLMs to store and retrieve contextual information across sessions.
Curated Intelligence
Users prompted ChatGPT with a photo of an empty room and the instruction to fill it using everything the model knows about them, and the output surfaced detailed personal facts, including an insect identified a year earlier. The viral test offers a vivid demonstration of the privacy risks posed by ChatGPT's long-term memory and cross-session data accumulation.
Silicon Valley startup AFK AI says its product Comma already implements a single-router, multi-worker multi-agent architecture, using continuous context and long-term memory to keep agents running 24/7 across many parallel tasks, with code open-sourced. The piece frames this as converging with DeepSeek Harness's Super-agent vision around human attention as the bottleneck, but offers mostly conceptual narrative without verifiable architecture details or benchmark results.
AFK AI says its product Comma already implements the single-Router, multi-Worker super-agent architecture later outlined by DeepSeek's Harness team, adding persistent memory, continual learning and 24/7 uptime to address the single-session context bottleneck. The system is publicly available with open-sourced code, and the piece frames human attention—capped at roughly 3-5 active sessions—as the real constraint on agent concurrency.
Honor CEO Li Jian told the 2026 Mobile Network Conference that competition over AI entry points is not a zero-sum game, arguing that scarcity in the intelligent economy has shifted from connectivity and attention to delegated intent, with monetization moving toward outcome-based billing. He outlined three barriers—intelligence, context and trust—and announced that YOYO will link agent partners including Lingxi, Qwen, Doubao and Gemini, plus a Token bundle launched with China Mobile.
SenseTime released SenseNova 6.8 Flash, centered on two capability directions: Office Intelligence, which maintains user goals, preferences and constraints across tasks and proactively recommends next actions from new context, and Dynamic Design, which produces animated PPTX decks, motion posters, interactive websites and editable Blender projects. The positioning shifts from isolated content generation toward task continuity and ready-to-use artifacts.
A new Tsinghua University paper reveals that AI agents with long-term memory fail to cleanly delete personal data upon user request. Testing across six memory systems showed an end-to-end success rate of less than 1.5%, with most deleted facts still extractable via simple prompts. The study urges developers to implement and verify actual deletion paths rather than relying on superficial confirmation messages.
SenseTime officially released the SenseNova 6.8 Flash model, focusing on enhanced office collaboration and design generation capabilities. The new version supports long-term memory and cross-session task coordination, featuring native generation for dynamic PPTs, SVGs, Blender 3D, and web designs, aiming to optimize visual storytelling and information hierarchy through agent collaboration.
Alibaba Cloud has upgraded its Memory service to enhance long-term retention of user preferences and history for Personal Agents, supporting integration via MCP and CLI. The service leads in HaluMem and PersonaMem benchmarks and allows developers to customize memory rules and combine them with Skills for reusable, personalized workflows.
A project featuring a Claude Opus 5.5-based agent has published its 'continuous life' notebook, demonstrating autonomous memory, literary analysis, and creative writing capabilities under a heartbeat wake-up mechanism. The project highlights experimental progress in long-context retention and autonomous behavioral patterns through detailed interpretations of 'The Tale of Genji' and short story generation.