
可能是最擅长「开会」的 AI 硬件。
AI 墨水屏赛道正在起势。
极客公园获悉,AI 硬件品牌 Flowtica 完成数千万人民币新一轮融资,由祥峰资本领投,源码资本、五源资本及清水湾基金等老股东跟投。与此同时,9 月在 IFA 宣发的 AI 墨水屏机型—— Verso,已正式在海外启动定金预售,并将于 12 月交付首批机型。
Verso 是一块 78 克、6.5 毫米厚的磁吸墨水屏,贴在手机背面。开会时按一下开始录音,实时转写、实时总结,结束录音时自动生成待办;过程中,单击标记重点,双击生成待办,在墨水屏上可以实时查看;按一下 Ask 键,可以问它任何关于你上下文的难题,还能直接给它交代一些任务。

右上方第一个按键,第二个是 AI 按键,一键记录灵感和待办,也可以 Ask AI|图片来源:Flowtica
Flowtica 联合创始人 Rian 告诉极客公园,Verso 的定位是「墨水屏工作伴侣」——福尔摩斯身边的华生,一个很懂使用者的辅助角色。他们聚焦的情形是,让使用者在一块墨水屏上,做会议前、中、后的完整闭环。
换言之,他们的目标可能是要成为最擅长「开会」的 AI 硬件。
Rian 毕业于香港科技大学,师从著名创业导师李泽湘教授,是一位兼具算法与硬件跨界背景的连续创业者。
他的创业履历几乎是按硬件品类排的:2016 年做出全球首款智能手机手持云台 Vimble,在 Kickstarter 众筹破百万,出货超 30 万台,后来被国内知名厂商总体收购;2017 年作为联合创始人做智能猫砂盆,用机器学习攻克多猫识别,海外众筹超 130 万美元;2018 切入农业科技,把高精度控制算法与计算机视觉用在农业水果分选上,做到行业第二。
从 2024 年起,Rian 重新回到 to C 领域,做 Flowtica。在最新的机型 Verso 上,他们给出的回答与这个时代的主流叙事恰好相反。在所有 AI 硬件都在讲「尽可能多收集上下文」的当下,他做了一个令投资人反复追问的决定:砍掉 Always-on 持续监听能力。
笔者与 Rian 聊了聊,手机已经足够强大、大模型能承载更多能力的当下,一块独立的墨水屏究竟凭什么存在?又如何建立护城河?
作者|苏子华
编辑|靖宇
为什么手机背后需要
一块「不发光」的卡片?
手机背面正在变得拥挤。AI 录音卡、墨水屏、AI 电子吧唧,都想占据那个位置。不过,为什么使用者还需要一块独立的墨水屏,而不是手机副屏、折叠屏,或者手表?
Rian 认为,如果只是一块彩色的 LCD 屏,充其量只是手机显示屏的延伸,注定会被手机的高刷与多元生态所吞噬。墨水屏的价值,在于它是一块反着来的显示屏——它可以一直在那儿,静静表明,不耗电。今天的下一件事是什么,随时一瞥就能看到。
但更本质的差异在于交互逻辑。
「语音是最高质的输入方式,但它不是高质的输出方式。」Rian 做过一个很具体的推演:如果 Agent 用语音给你回一段 50 字的内容,可能需要 30 秒才能听完;同样一段信息展现在显示屏上,你扫一眼就完成了。语音的线性特征决定了它无法被快速浏览、无法被跳读、无法被结构化展现。
「笔者把硬件当成一个被完整包装了 UI 与物理交互的实体 APP。它就是使用者的一个 APP,随时可以打开录音、查看日程和 todo,直达性比手机慌忙解锁、点半天强得多。」
在极高时薪与高压的知识工作情形中,使用者最宝贵的是「心流(Flow)」。相比之下,传统手机的交互路径是破坏性的:为了看一眼下一个日程,你必须拿起手机、解锁显示屏、穿过推送通知的狂轰滥炸,最终极易迷失在诸如小红书、抖音、微信等社交软件与短视频的信息海洋中。
「手机本质上是处理复杂算力与通讯的超级终端,但高沉浸度的生产力工作需要专有的物理载体。 」在 Rian 看来,Verso 不是手机的替代品,而是一个低打扰的视觉反馈面板,和一个随时掌控全局的控制型小设备。

Rian 认为,手机生态本身正在被动摇:AI 能力变强后,手机的许多传统履约正在被 Agent 接走。但从生态位的博弈来看,Verso 避开了与智能手机在「万能终端」上的正面硬刚。
「笔者在做的事情没有跳开现有生态,和 Apple 体系、安卓体系还是好朋友:连日历、连既定的习惯平台。同时,做格外垂直、增量的事情。」
Flowtica 也评估过做耳机、手表的形态,但最终放弃。理由是,耳机带宽太窄,派给 AI 的活需要时间完成,做得怎么样,需要看见。「显示屏给人安全感,因为它就在那儿,也方便结构化展示。」
至于手表,致命缺陷是显示屏小,视觉信息带宽过狭,无法给出视觉上结构化总结的安全感。
聚焦办公:把会议的
前、中、后做成闭环
Verso 的定位是「墨水屏工作伴侣」(ePaper sidekick)。在构建 AI 硬件上,团队的思路是精准地聚焦,集中在高频、高压力的办公会议情形。
Rian 表示,团队坚信,对于知识工作者而言,会议绝对不是漫无目的的聊天,而是绝大多数实际工作的起点与决策枢纽,「很多高价值 context 发生在沟通上」。
这个判断已经被 Flowtica 第一款机型 AI 录音笔 Scribe 上验证过——在使用者首次上手引导的情形挑选中,「内部会议」与「客户会议」稳居前两位;而 Scribe 录音的平均时长接近 40 分钟,几乎就是一场标准会议的长度。在他们看来,这是最可以撬动使用者付费意愿的情形。
因而,他们将 Verso 优先嵌入到会议的前、中、后全生命周期。
在会前,依托对使用者日历与历史交互记忆的深度绑定,Verso 能在会议开始前主动展现在墨水屏锁屏界面。它会依据上一次与该对象的沟通,自动生成一份极简的会前 Brief,协助使用者无缝进入谈判或决策状态。
在会中,面对面交流或电话会议中,按下录音键,即可启动高清录音与实时转写。同时,单击可「标记重点」,双击可「生成 Todo」,配合轻微的物理震动反馈——从第一代 Scribe 开始,它是 Memory 最早的主动输入,也是 AI 理解使用者的第一步。
在会后,Verso 会提取出高度结构化的核心决议,无缝导出行动项或跟进邮件草稿。这些 Todo 实时同步到墨水屏看板上,方便使用者像消杀游戏关卡一样在小方块里逐一清空任务。

一天结束,它会告诉你「你今天达成了什么」。所有笔记、会议、todo、日历,都是 knowledge base 的一部分,也是后续 Agent 交付的素材。
另外,Verso 还集成了个人日历、邮箱、灵感摘要、电子书阅读以及基于 NFC 的数字名片能力。
反行业共识:
为什么放弃 Always-on
在今天的创投语境里,Always-on 持续监听录音几乎是 AI 硬件的标配叙事:上下文越多越好,Agent 越懂你。Verso 是反着来的。
Rian 表示,在 2024 年,他们就提出了 Always-on 概念,「从工艺上讲,做到 Always-on 没有难度」。不过,做着做着,他们自己发现,在目标人群里这可能不是正确答案。
团队调研后发现,首先,使用者不喜欢上下文被腐蚀。「高时薪人群不希望工作外的 context 混进来,那对他是上下文腐蚀,不是增强。」
第二,海外使用者对隐私极度敏感,Always-on 在很多地区甚至不可用。欧盟对苹果的 Siri AI、对 Amazon 门铃都有限制。「它不仅是能力难题……还有隐私和监管难题。」
Flowtica 团队在测试中发现,当使用者带着设备从高密度的商业会议情形,走到路人八卦的咖啡馆或私密的家庭情形,如果一股脑将所有无序的无意义对话塞进大模型的记忆上下文,整个 Agent 的记忆平台反而会走向崩塌,输出大量充斥着噪音的废话。
在 Rian 看来,这是一次典型的工艺 ready 了,但消费者没 ready 的取舍。
「对 Always-on 觉得格外渴望的人,是金字塔尖,巴不得人机共生。但对更大的人群,仍然要监管、习惯、社会接受度、心理接受度都融合,事情才能往下走。」他也承认,5 年后接受 Always-on 的人超过 60% 完全可能,「那时候人们感受过 context 越多、Agent 服务越好。但当下是强烈冲突」。
因而,Flowtica 的机型当下仍然需要手动开关录音键。
下一个高地
Verso 不是 Flowtica 推出的第一款机型。2024 年底,Flowtica 同时前测了两款机型:一支录音笔 Scribe,一块带墨水屏的卡片 Verso。团队最终决定分步走,先做笔,当时还没有 Agent。

Flowtica 的 AI 录音笔 Scribe|图片来源:Flowtica
「Agent 没来时,跟人们说 Agent-native 硬件,人们没感觉;Agent 来了,有屏、可直接语音交互、带着 context、随时懂你、可以派活的硬件,格外 make sense。」Rian 回忆道,2025 年底,Verso 迎来了真正的启动契机。
Flowtica 团队有自己的审美坚持,甚至花费高额成本重新为墨水屏定制了专属字体,调校出极具工艺挑战的「非对称」外壳与侧边盲操按键。
「在笔者团队,好看是底线。做出不好看、不好用的机型,人们会产生生理上的难受。」
不过,从 AI 录音卡到 AI 墨水屏,这个品类正在肉眼可见地同质化。Rian 对此很清楚,单纯的「语音转写」与「文字总结」正在迅速失去溢价空间。
他认为,下一阶段的竞争高地,在于 Agent Memory 与交付物的生产,「录下来的东西怎么变成接下来有用的价值、怎么机型化,很关键。」

Flowtica 在构建的多端生态,「个人生产力硬件」的蓝图|图片来源:Flowtica
因而,Flowtica 正在构建深度的 Context Engineering 与多端联动平台,让 AI 可以穿透时间。在跨时间线串联维度,它可以自动关联过去半年内与同一客户、同一项目发生的数次会议,准确提取出心态演变与项目发展脉络,让 AI 真正摆脱「幻觉」。
在实质性工作交付维度,结合 Flowtica 在 Mac 端、Watch 端以及 Web 端的软硬一体布局,Agent 可以依据沟通沉淀,直接帮使用者生成 HTML 报告、PPT 等。「沟通的结束,才是真正工作的开始。谁能帮使用者把会后的事情真正推进下去、拿到商业结果,谁才能享有极高的溢价。」Rian 强调。
在拥挤的赛道里,这块墨水屏真正的可能性,也才刚刚展开。
“特别声明:以上作品内容(涵盖在内的视频、图片或音频)为凤凰网旗下自媒体平台“大风号”使用者上传并发布,本平台仅给出信息存储空间服务。
Notice: The content above (including the videos, pictures and audios if any) is uploaded and posted by the user of Dafeng Hao, which is a social media platform and merely provides information storage space services.”
本文来源:https://tech.ifeng.com/c/8xNKug2OH8D
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。
Probably the best AI hardware for “meeting”.
Geek Park was informed that AI hardware brand Flowtica has completed a new round of financing of tens of millions of yuan, led by Xiangfeng Capital, and followed by existing shareholders such as Source Code Capital, Wuyuan Capital and Qingshuiwan Fund. At the same time, Verso, an AI ink screen model announced at IFA in September, has officially launched a deposit pre-sale overseas and will deliver the first models in December.
Verso is a 78g, 6.5mm thick magnetically inked screen attached to the back of the phone. During the meeting, click Start Recording, transcribe in real time, summarize in real time, and automatically generate a to-do when you end the recording. In the process, click the highlight, double-click to generate a to-do, and you can view it in real time on the ink screen. Click the Ask key to ask it any difficult questions about your context, and you can directly explain some tasks to it.
The first button in the upper right, the second is the AI button, which is a one-click record of inspiration and to-do, or you can ask AI | Image source: Flowtica
Rian, co-founder of Flowtica, told Geek Park that Verso is positioned as an “ink-screen work companion” - Watson next to Sherlock Holmes, an auxiliary who understands the user. They focus on having the user do a complete closed loop before, during, and after the meeting on an ink screen.
In other words, their goal may be to be the AI hardware that is best at “meeting.”
Rian graduated from the Hong Kong University of Science and Technology and studied under Professor Li Zexiang, a renowned entrepreneurial mentor, and is a serial entrepreneur with a cross-border background in algorithms and hardware.
His entrepreneurial career is almost ranked by hardware category: in 2016, he made the world's first smartphone handheld gimbal Vimble, broke millions in Kickstarter crowdfunding, shipped more than 300,000 units, and was later acquired by well-known domestic manufacturers; in 2017, he was a co-founder of the smart cat litter box, using machine learning to overcome multi-cat identification, and raised more than 1.3 million US dollars overseas; in 2018, he cut into agricultural technology, and used high-precision control algorithms and computer vision in agricultural fruit sorting, making it the second largest in the industry.
From 2024, Rian returns to the to C domain as Flowtica. On the latest model, Verso, they gave the opposite answer to the mainstream narrative of the era. At a time when all AI hardware is talking about "collecting as much context as possible", he made a decision that investors repeatedly questioned: cutting Always-on continuous listening capability.

I chatted with Rian. Now that the mobile phone is powerful enough and the large model can carry more capacity, why does a separate ink screen exist? How do you build a moat?
The back of the phone is getting crowded. AI recording cards, ink screens, AI electronic bars, all want to occupy that position. But why would a user need a separate ink screen instead of a phone sub-screen, a folding screen, or a watch?
Rian believes that if it is only a color LCD screen, it is only an extension of the mobile phone display at best, and is destined to be swallowed by the high brush and diverse ecology of the mobile phone. The value of the ink screen is that it is a reversed display - it can be there all the time, quietly showing that it does not consume electricity. What's next for today can be seen at a glance.
“Voice is the highest quality input, but it's not the highest quality output." Rian made a very specific deduction: If the Agent returns a 50-word piece of content in voice, it may take 30 seconds to finish listening; the same piece of information is displayed on the display, and you can finish it at a glance. The linear nature of speech determines that it cannot be quickly navigated, skipped, or structured.
"The author regards the hardware as a fully packaged physical app that interacts with the UI and physics. It is a user's app that can open the recording, view the schedule and todo at any time, and the directness is much stronger than the mobile phone's hasty unlocking and half a day."
In extremely high hourly pay and high-pressure knowledge work situations, the most valuable thing for users is "flow". In contrast, the interaction path of traditional mobile phones is disruptive: in order to see the next agenda, you must pick up the phone, unlock the display, go through the bombardment of push notifications, and eventually easily get lost in the information ocean of social software such as Xiaohongshu, TikTok, WeChat and short videos.
“Mobile phones are essentially super-terminals that handle complex hashrate and communications, but highly immersive productivity work requires proprietary physical carriers." In Rian's view, Verso is not a replacement for a mobile phone, but a low-disturbance visual feedback panel and a small control device that can control the whole picture at any time.
Rian believes that the mobile phone ecosystem itself is being shaken: many of the traditional fulfillments of mobile phones are being taken away by agents after AI capabilities become stronger. But from the perspective of the game of ecological niche, Verso avoids the front and back of the "universal terminal" with smartphones.
"What I'm doing is not skipping the existing ecology, and I am still good friends with the Apple system and the Android system: even the calendar, even the established habit platform. At the same time, do extra vertical and incremental things.”
Flowtica also evaluated the shape of the headphones and watches, but eventually gave up. The reason is that the bandwidth of the headphones is too narrow, and the work assigned to the AI needs time to be completed, and how it is done needs to be seen. “The display feels safe because it's there and it's structured.”
As for watches, the fatal flaw is that the display screen is small and the bandwidth of visual information is too narrow to give a sense of security for visually structured summaries.
Verso is positioned as an ePaper sidekick. In building AI hardware, the team's idea is to focus precisely on high-frequency, high-stress office meeting situations.

Rian said that the team firmly believes that for knowledge workers, meetings are definitely not aimless chats, but the starting point and decision-making hub for most practical work, and “a lot of high-value contexts happen in communication”.
This judgment has been verified on the Scribe of Flowtica's first AI recorder - the "internal meeting" and "customer meeting" ranked in the top two in the selection of the user's first onboarding situation; while the average length of the Scribe recording is close to 40 minutes, which is almost the length of a standard meeting. In their view, this is the best case scenario to leverage users' willingness to pay.
As a result, they prioritize embedding Verso into the entire life cycle before, during, and after a meeting.
Before the meeting, Verso can proactively display the ink screen lock screen before the meeting starts, relying on the deep binding of the user's calendar and historical interactive memory. It will automatically generate a minimal pre-session brief based on the last communication with the object to assist the user to seamlessly enter the negotiation or decision-making state.
In the meeting, face-to-face communication or conference call, press the recording key to start high-definition recording and real-time transcription. At the same time, click to “Mark Focus” and double-click to “Generate Todo” with slight physical vibration feedback - starting with the first generation of Scribe, it is the earliest active input of Memory and the first step for AI to understand users.
After the meeting, Verso pulls out highly structured core resolutions, seamlessly exporting action items or follow-up draft emails. These Todos are synchronized to the ink screen board in real time, making it easy for users to clear tasks one by one in small squares like killing game levels.
At the end of the day, it tells you “What did you achieve today?” All notes, meetings, todo, calendars, are part of the knowledge base and are the material for subsequent Agent deliveries.
Verso also integrates personal calendars, email, inspiration summaries, e-book reading, and NFC-based digital business card capabilities.
In today's venture context, Always-on continuous monitoring recording is almost a standard AI hardware narrative: the more context, the better Agent understands you. Verso is the opposite.
In 2024, Rian says, they came up with the concept of Always-on, “technically, there's no difficulty in being Always-on.” However, while doing it, they themselves found that this may not be the right answer in the target group.
The team investigated and found that, first of all, the user did not like the context to be corrupted. “The high hourly wage crowd doesn't want the outside context mixed in, it corrupts the context for him, not enhances it.”
Second, overseas users are extremely sensitive to privacy, and Always-on is not even available in many regions. The EU has restrictions on Apple's Siri AI, and on Amazon doorbells. “It's not just a competency challenge… it's also a privacy and regulatory challenge.”
In the test, the Flowtica team found that when the user walked from a high-density business meeting situation to a passerby's gossip cafe or a private family situation with the device, if a brain crammed all the disordered and meaningless conversations into the memory context of the big model, the entire Agent's memory platform would collapse and output a lot of noise-filled nonsense.
In Rian's view, this is a typical process ready, but consumers are not ready to make trade-offs.
“For those who feel particularly eager for Always-on, it is the tip of the pyramid that desires human-computer symbiosis. But for the larger population, supervision, habits, social acceptance, and psychological acceptance must still be integrated before things can go down. He also admitted that it is entirely possible that more than 60% of people will accept Always-on in 5 years.“ The more people feel the context, the better the Agent service. But right now it's a fierce conflict. ”
Therefore, the Flowtica model still needs to manually switch the recording button for the time being.
Verso is not Flowtica's first model. At the end of 2024, Flowtica tested two models at the same time: a recording pen Scribe and a card Verso with an ink screen. The team finally decided to go step by step and make a pen first, there was no agent at that time.
Flowtica's AI Recorder Scribe | Photo Credit: Flowtica

"When the Agent did not come, people told people about the Agent-native hardware, people did not feel it; the Agent came, with a screen, direct voice interaction, with context, hardware that understands you at any time, and can be dispatched to make sense. Rian recalls that at the end of 2025, Verso had a real launch opportunity.
The Flowtica team has its own aesthetic insistence, and even spent a lot of money to re-customize the exclusive font for the ink screen, adjusting the "asymmetric" shell and side blind button with great process challenges.
"In the author's team, good looks are the bottom line. Making a model that is not good-looking and not easy to use will cause people to feel physiologically uncomfortable.”
However, from AI recording cards to AI ink screens, this category is becoming visibly homogenized. Rian is well aware that simple “speech transcription” and “text summaries” are rapidly losing their premium space.
He believes that the next stage of competition is in the production of Agent Memory and deliverables. "How the recorded things become the next useful value and how to model is very important."
Flowtica is building a multi-faceted ecosystem, a blueprint for "personal productivity hardware" | Image source: Flowtica
As a result, Flowtica is building a deep Context Engineering and multi-end linkage platform that allows AI to penetrate time. In the dimension of tandem across timelines, it can automatically correlate several meetings with the same customer and the same project in the past six months, accurately extract the evolution of mentality and project development, so that AI can truly get rid of "hallucinations".
In the substantive work delivery dimension, combined with Flowtica's software and hardware layout on the Mac side, Watch side, and Web side, the Agent can directly help users generate HTML reports, PPTs, etc. based on communication precipitation. “The end of communication is the beginning of real work. Who can help users really push things forward after the meeting and get business results, who can enjoy a very high premium.” Rian emphasized.
In a crowded circuit, the real possibilities of this ink screen are just beginning to unfold.
"Special statement: The contents of the above works (including videos, pictures or audio) are uploaded and published by the users of the media platform" Dafeng "owned by Phoenix Networks, and this platform only provides information storage space services.
Notice: The content above (including the videos, pictures and audios if any) is uploaded and posted by the user of Dafeng Hao, which is a social media platform and merely provides information storage space services. "
This article is from: https://tech.ifeng.com/c/8xNKug2OH8D For copyright issues, please write to: 13951431913#139.com or submit a review on the right.

