智能硬件的竞争正在从参数转向记忆能力。10月24日,红熊AI在WTCC 2026西岸国际科技消费嘉年华发布智能录音卡笔记熊。这款机型由一块录音卡、一个手机App和一套云端记忆引擎构成,搭载红熊AI自研的类人脑记忆引擎MemoryBear。

红熊AI主攻AI记忆科学,笔记熊是这套能力在消费级硬件上的载体,定位“个人第二大脑”。它的任务不止于转写:把会议里的决定、聊天中的细节、一闪而过的想法真正记住、理清,在需要的时候找出来,变成下一步工作的起点。
从录音转写工具到记忆伙伴,笔记熊靠的是一套记忆引擎
AI工艺进入日常工作后,常常卡在失忆这个瓶颈上:对话内容、使用者偏好、上下文在会话结束时就断了,同一个难题隔一段时间再问,模型接不上前因后果。红熊AI把神经科学、认知科学与计算机科学融合在一起,参照人脑海马体的记忆机制自研类人脑记忆引擎MemoryBear,用记忆科学破解AI失忆难题。
过去这套引擎以软件形态嵌入企业与第三方平台,涵盖大模型、CRM、ERP;装进录音卡之后,同一套引擎在消费级硬件上跑通。
笔记熊的三项能力,覆盖记忆的组织、调用与更新
记忆图谱:录音内容被拆成条目,按人、项目和当时定下的结论关联成一张网,不再只是按文件堆叠。
随问随答:使用者用自然语言提问,平台从数月前的录音里检索答案并标出来源,说明出自哪场会议、哪一段录音;查不到出处时平台会直接说明,不做推测。
记忆更新:新信息进来会改写旧结论,记忆随使用一起生长。
以机型评审会为例,会议结束纪要同步生成:结论排在前面,每个议题下分列决策、理由与后续动作,参会人和负责人自动标注,行动项推到录音卡卡面。销售情形一样有用,每次拜访同一客户前打开“会前复习”,上次聊到哪、预算有没有改、哪些事还没办完一次看全,此前的录音不必重听。

分层记忆存储,个人记忆默认仅本人可见
在使用者数据上,红熊AI分三层处理:个人、团队、企业各自独立存储,个人记忆只有本人能看,并能在使用者授权下达成团队、企业共享或脱敏后共享。
红熊AI模型研发总监孙科在发布会现场表示:“记忆不是附加能力,它决定AI能不能理解一个人、一个团队正在做的事。借助记忆能力,笔记熊把过去散落在会议、对话和想法里的上下文留存下来,让记忆有结构、有来源、有时间、会演化,在需要的时候能找到、能用上。”

从笔记熊出发,红熊AI启动“记忆科学+智能硬件”的生态布局。笔记熊后续会与记忆型全模态通用大模型OpenBear、记忆型编程大模型CodeBear等机型打通,多个Agent可以基于授权的记忆分工协作。环境感知、智能家居、陪伴玩具、时空感知等形态也在延伸范围内,这几类设备持续产生数据,却缺少把数据沉淀成长期记忆的能力。
关于红熊AI:公司主攻AI记忆科学,自研记忆型大模型与类人脑记忆引擎MemoryBear,使命是“把人工智能带入每一家企业”。公司2026年7月完成A+轮融资,金额数亿元,投后估值近30亿元,投资方涵盖浙江九纬私募基金、嘉兴彰元创投与老股东格睿丰,这是其15个月内完成的第6轮融资;此前获长城战略咨询“中国潜在独角兽企业”“中国哪吒企业”认定。
本文来源:https://www.bjnews.com.cn/detail/1792832260129490.html
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
Competition in smart hardware is shifting from specs to memory capabilities. On October 24, Hongxiong AI unveiled the Smart Voice-Card NoteBear at the WTCC 2026 West Bund International Tech Consumer Festival. The system pairs a voice-recording card with a mobile app and a cloud-based memory engine, powered by Hongxiong AI’s proprietary human-brain-inspired MemoryBear engine.
Red Bear AI specializes in AI memory science, with Note Bear serving as the consumer hardware platform for this technology, positioned as a "personal second brain." Its mission goes beyond mere transcription: it truly captures and organizes meeting decisions, chat details, and fleeting thoughts, retrieving them when needed to become the starting point for your next task.
From audio transcription tool to memory partner, Note Bear relies on a memory engine.
As AI processes become integrated into daily work, they often hit a bottleneck: forgetfulness. Conversation content, user preferences, and context disappear when a session ends, so if the same difficult question is posed again after some time, the model cannot pick up where it left off. To tackle this, Hongxiong AI fuses neuroscience, cognitive science, and computer science, and—drawing on the memory mechanisms of the human hippocampus—has independently developed a human-brain-inspired memory engine called MemoryBear, applying memory science to solve AI’s amnesia problem.
Previously, this engine was embedded as software in enterprises and third-party platforms, covering large models, CRM, and ERP systems. After being integrated into recording cards, the same engine now runs successfully on consumer-grade hardware.
Note Bear's three capabilities cover the organization, retrieval, and updating of memory.
Memory Graph: Recordings are broken down into entries and interconnected into a web based on people, projects, and decisions, rather than just being stacked in files.
Ask and Answer: Users pose questions in natural language, and the platform retrieves answers from months of recordings, citing the source by meeting and specific segment. If no source is found, the platform explicitly states so rather than speculating.
Memory update: New information comes in and rewrites old conclusions; memory grows alongside use.
Take design review meetings as an example: meeting minutes are generated simultaneously, with conclusions listed at the top. Each agenda item details the decision, rationale, and next steps, while attendees and owners are auto-tagged, and action items are pushed directly to the recording card. This is equally useful for sales. Before visiting a client, simply open the "Pre-meeting Review" to see your last discussion, any budget changes, and outstanding tasks at a glance, eliminating the need to re-listen to previous recordings.
Tiered memory storage; personal memories are visible only to the user by default.
Regarding user data, Hongxiong AI utilizes a three-tier processing system where individual, team, and enterprise data are stored independently. Personal memory data remains visible only to the individual user; however, with user authorization, it can be shared with teams or the enterprise, or shared after being desensitized.
Sun Ke, Director of AI Model R&D at Red Bear AI, stated at the launch event: "Memory is not just an add-on capability; it determines whether AI can truly understand what an individual or team is doing. Leveraging this capability, Note Bear preserves context previously scattered across meetings, conversations, and ideas. It ensures memory has structure, source, and timeline, and evolves, making it retrievable and applicable when needed.
Starting with the Note Bear, Red Bear AI is launching an ecosystem built on "memory science + smart hardware." The Note Bear will subsequently integrate with models such as the memory-based omni-modal general large model OpenBear and the memory-based coding large model CodeBear, enabling multiple agents to collaborate by dividing tasks based on authorized memories. Forms such as environmental sensing, smart home devices, companion toys, and spatiotemporal perception are also being explored; while these devices continuously generate data, they lack the capability to transform that data into long-term memory.
Red Bear AI specializes in AI memory science, having self-developed a memory-based large model and a human-brain-like memory engine called MemoryBear. The company's mission is "to bring artificial intelligence into every enterprise." In July 2026, it completed a Series A+ funding round, raising hundreds of millions of yuan and reaching a post-money valuation of nearly 3 billion yuan. The investors included Zhejiang Jiuwei Private Equity Fund, Jiaxing Zhangyuan Venture Capital, and existing shareholder G-ridge. This marked the company's sixth funding round in just 15 months. Previously, it was recognized by the Great Wall Enterprise Institute as a "China's Potential Unicorn Enterprise" and a "China's Nezha Enterprise.
Source: https://www.bjnews.com.cn/detail/1792832260129490.html. If there are any copyright issues, please contact us at 13951431913@139.com or submit a comment on the right.