第一款产品做“AI办公助理”，一个尚未达成共识的产品形态。
独家获悉，前网易副总裁、杭州研究院执行院长、网易数帆总经理汪源近期成立了杭州久痕科技公司并担任CEO，第一款产品的定位是做“以个人数据为中心的通用办公助理”。
久痕科技的愿景是“成为世界级的人工智能企业”，使命是“让知识工作更轻松高效”。
汪源是一名网易老兵，在网易将近18年。汪源参与领导的网易杭州研究院成功孵化了网易云音乐、网易云课堂、网易严选、Lofter、网易云阅读、网易博客、网易相册等众多C端产品和网易数帆、网易易盾、网易轻舟、网易CodeWave等众多B端业务，是国内唯一取得持续多次成功的企业研究院机构（类似的机构有曾经的盛大创新院和创新工厂）。 数字化转型网（www.szhzxw.cn）
久痕科技的另一位核心高管为陈成丰，担任久痕科技工程研发负责人。陈成丰曾是杭州出海头部企业小影创新科技联合创始人，担任小影创新技术VP。在创立小影科技之前，陈成丰曾担任虹软Mobile BU技术总监，MediaPlayer与Camcorder负责人，负责韩国三星、LG等手机厂商的多媒体技术OEM开发与交付。
汪源在大约4个月前决定离开网易创业。今年5月22日，汪源在自己的公众号《冷技术热思考》中官宣了从网易离职创业的消息。当时汪源表示会回到自己熟悉的to C市场。他判断随着大模型和硬件供给能力的成熟，生成式AI应用在今明两年会迎来新的窗口期，特别是在欧美市场。
汪源创业的消息吸引了数百名投资人的关注。目前久痕科技已经完成种子轮融资，主要是美元基金，为初始团队储备了两到三年的充足资金。公司正在招聘产品合伙人与算法工程师。
久痕科技的第一款产品定位是“以个人数据为中心的通用办公助理”，就像秘书一样帮助个人完成参加会议、回复消息、回复邮件、阅读文档、写作报告、管理日程等各类典型的知识工作。
汪源判断，通用办公助理市场空间巨大。面向全球十亿规模的知识工作者，按照5-10%的付费用户比例计算，存在数千万甚至亿级的付费用户规模，按100-200美元/年的收费标准，存在数十亿乃至百亿美元的营收规模，对应公司市值可达千亿美元。
但个人办公助理的产品形态究竟长什么样子，目前也很难定义清楚，不少投资人并不看好这一方向。比如，金沙江创投主管合伙人朱啸虎今年3月在接受采访时提出一个“暴论”：“今天说‘个人助理’都是技术人员的想象。我就问你，有几个人需要个人助理？都是典型的伪需求！”
汪源对此表示理解。他认为很多人难以想象这个产品的形态，这是好事，因为伟大的产品总是来自于一小群人的共识，不可能来自于大多数人的共识。 数字化转型网（www.szhzxw.cn）
市场上还没有久痕科技完全对标的产品，唯一稍有类似的Rewind。但久痕科技想做的产品，包含了五倍于Rewind的个人数据——Rewind只能通过录屏录音的方式慢慢积累数据，以及十倍百倍于Rewind的功能性——Rewind几乎只做了关键字搜索功能。
因为涉及个人办公数据处理，必须确保数据运行在用户本地的端侧大模型，这就必须依赖于AI PC市场的成熟。显然，AI PC市场仍在早期，汪源形容今天的AI PC类似于2010年iPhone4刚发布，因此现在是创业入局的好时机。
在第一款产品之后，久痕科技计划“以个人数据为中心”为基石推出更多办公领域的产品，如代办管理、笔记等，力争成为“AI时代的Office”。
跟大多数AI to C的产品一样，久痕科技从第一天起就进行全球布局，核心面向欧美日为主的海外市场，采取订阅制收费。 数字化转型网（www.szhzxw.cn）
公司名「久痕」来自于乔布斯的名言 make a dent in the universe。汪源表示，乔布斯领导的Apple公司在PC、智能手机、平板电脑、在线音乐、TWS耳机等领域都开创或普及了新的产品品类，成功留下痕迹。这次创立久痕科技，汪源希望致敬乔布斯和延续网易杭州研究院的创新精神，在业界留下更多痕迹。
最近，有记者专访了汪源。关于AI创业，他是怎么思考的？
一、为什么创业？
Q：创业这件事，你考虑了多久？
汪源：其实我在2019年就希望自己能出来做点事情，但是在网易我管理的各部门业务，确实也蛮难去交接的，老板也一直在给我更多的职责。那时候如果我走，对老板也不好交代。所以就拖了四五年的时间。
Q：你在自己的公众号里宣布创业，说是“事了拂衣去看世界”，“事了”是什么意思？
汪源：我在网易负责两大块业务，第一块是后端如数据中心与IT体系建设。IT体系去年成功实现了全集团的整合，在贵州的大规模自建数据中心也成功投产了。
第二块是近几年我最核心的工作，负责管理网易数帆在企业服务领域的大数据、云原生、低代码、to G等相关业务。去年我基本上把业务相应的部门建设、业务发展做到了还不错的状态。相比几年前从零开始，如今已经做到了业务成型，部分业务跻身一线供应商的水平了。所以，这一块的业务也告一段落。
总的来说，我担任总经理的网易数帆和智企整合成新的数智事业部，我负责的杭州研究院的其他工作也都有了较为妥善的交接和安排。 数字化转型网（www.szhzxw.cn）
Q：有考虑在网易内部做这件事吗？
汪源：我没有在内部提过。第一个原因是，我认为公司最近几年的战略还是非常聚焦的，基本上不会去开新的业务线。第二，网易内部有很大的存量业务需要我去管理，我肯定也没有足够的精力去拉起一条在业内都没有出现过的产品形态，这个难度需要个人投入非常大的精力去做。
我了解到现在很多的AI创业公司都是模仿美国产品，比如Gamma火了就有很多AI PPT产品。模仿当然也很好，但我们没有一个完全对标的产品，而且会直接到美国市场去做从0到1的创新。这个工作的难度与挑战需要我全职全身心的投入。
Q：过去几年你最核心的业务是网易数帆，是一个to B业务，但为什么创业却选择做to C？
汪源：其实我在网易18年的时间段里，参与时间最长的还是to C业务。除了网易数帆to B之外，网易云音乐、网易云课堂等产品都是to C的。第二个考量是因为我们想做全球市场，to C会比较好启动。在美国做to B蛮难的，组建线下的销售队伍，管理肯定比to C难。 数字化转型网（www.szhzxw.cn）
二、为什么做AI办公助手？
Q：创业方向是如何考虑的？
汪源：我们决定做一款C端的应用层产品，而不是做foundation model，或者垂直行业大模型，比如医疗等。我去年做过将近一年的医疗大模型，但我感觉垂直赛道挺难发展，最后都变成了to B的项目定制化。
Q：你觉得to C市场有哪些不错的产品？
汪源：主要的产品有三类，第一类是情感陪伴、聊天类，体现大模型的情绪价值；第二类，如果你看a16z Top 50的报告，更多的是在图像和视频的内容编辑与内容生成领域，50个应用可能占一半；第三类是a16z最新报告里出现的一类，生产力产品。
生产力产品赛道已经出现了不少产品，比如国内的浏览器插件Monica、AiPPT等。我们属于生产力赛道的办公助手类产品。 数字化转型网（www.szhzxw.cn）
Q：为什么第一款产品的定位是以个人数据为中心的通用办公助手？
汪源：大模型有强大的通用知识和智能处理能力，但ChatGPT日均不到一次的使用数据说明它就像一个入职第一天的新手一样做不了多少具体工作，除非你的工作主要是写写套话。即便下一代大模型再强大，不了解用户，就不可能在大多数场景中有效的提供帮助。大模型会成为越来越标准化的商品，而用户数据会越来越宝贵。在办公助理领域，善于获取、组织和利用用户数据者得天下。
我们的产品就像一个无时无刻陪伴你多年的秘书，非常了解你的知识、思路和习惯，从而能够高效的帮助你完成参加会议、回复消息、回复邮件、阅读文档、写作报告、管理日程等各类典型的知识工作。
Q：飞书与钉钉也在做类似的办公助手。
汪源：不一样。飞书与钉钉实际上是特定的办公软件，而我们要做的办公助手本身不做办公软件，但它会像一个人一样去使用办公软件。我认为办公软件主体是功能性的，而非智能性的。我们的重点是做智能性产品，比如我帮你回复邮件，但我不会去做一个邮件客户端；我帮你写笔记，但不会做一个笔记软件。
另一方面我们主要面向海外市场。其实在海外市场，钉钉与飞书这种全家桶式的软件不是很受市场欢迎，大家倾向于用多个更加垂直的、更加专业的软件来完成工作。
Q：为什么首先选择美国市场？跟用户付费习惯有关系？
汪源：总的来说美国市场的体量和付费用户的意愿是最高的。我是做一个终端的、面向to C市场的软件，只涉及个人数据，所以并没有必要用to B的方式去做，to C的商业闭环也更加简单，付费链路也更短。美国市场肯定是全球最大的一个市场。
另外，美国市场对于软件的反盗版意识是比较强的。我们作为一个完全运行在PC上的软件，从技术上来说很容易被盗版，因为它不依赖于云服务。美国市场通常是比较关心这一点的，尤其是对于办公类工具，一般的美国公司不允许员工使用盗版软件。
我们的想法是先在美国市场做，等产品做出来了也看看国内市场能做多少，但国内确实没有一个很高的预期。 数字化转型网（www.szhzxw.cn）
Q：你们有没有对标产品？
汪源：有一个稍微类似的叫Rewind。Rewind最早是在PC上把你的屏幕状态和麦克风听到的声音实时录制。当你想起来前两天看到的一个内容，Rewind会帮你找到，就是这样一个简单的功能。
我的理解是，录屏只是个人数据的一小部分，比如你的电脑上已经有非常多的文档，浏览器里也打开过很多网页和文档，这些数据都没有抓取。另外Rewind只做了一个关键字搜索的功能，但没有做任何其他的功能，比如帮你写邮件、回消息、管理日程等。所以这个产品在火了一阵子之后，增长其实也不是特别好。我自己也是Rewind非常早期的用户，体验下来感觉对我的帮助也是非常低频的，可能几天才会用一次。这样的产品付费也不会很好。
Q：听起来对于数据安全会是一个极大的考验。
汪源：这也是我们认为现阶段有条件去做这件事的一个核心原因。理论上讲，我希望产品尽可能访问用户所有的办公资料，就必须确保处理这些数据的大模型必须在用户本地的端侧大模型，绝对不能采用云服务的方式来提供。 数字化转型网（www.szhzxw.cn）
因此，这个产品必须依赖于AI PC，当AI PC出现，我们的产品才可能出现。而AI PC现在还在早期，有点像2010年iPhone4刚发布的时代。这是一个创业的好时机。
Q：有多少投资人认可你做的产品方向？比如朱啸虎就说过，“AI个人助理是个典型的伪需求，是技术人员自己的想象”。
汪源：我没有跟朱啸虎聊过。确实有投资人最喜欢的逻辑就是问你这个产品对标的是美国哪个产品，但我们想做的是一个全新的产品形态。如果投资人脑子里没有一个具象的产品，他可能就不太确定这个方向到底是行还是不行，就会有顾虑，我觉得可以理解。在一个新的品类出来之前，大家其实都这样。说实话，当时iPhone出来后，还有很多人说iPhone也没什么用啊，不就是打电话、发短信吗？
Q：融资现在是什么情况？
汪源：我跟资方协议有保密条款，只能说都是美元基金，公司的整个架构也是面向海外市场，尤其是可能到美国去上市。但公司的产研团队是在国内，中国还是有工程师红利在的。
翻译：
Wang Yuan, former vice president of NetEase, established Long trace technology, and the seed round of financing has reserved sufficient funds for two to three years
The first product is “AI office assistant”, a product form that has not yet reached a consensus.
It was exclusively learned that Wang Yuan, former vice president of NetEase, executive director of Hangzhou Research Institute and general manager of NetEase Sufan, recently established Hangzhou Jiuchen Technology Company and served as CEO, and the first product was positioned to be a “general office assistant centered on personal data”. 数字化转型网（www.szhzxw.cn）
The vision of Jiuchen Technology is to “become a world-class artificial intelligence enterprise”, and the mission is to “make knowledge work easier and more efficient”.
Wang Yuan is a veteran of NetEase, nearly 18 years in NetEase. The NetEase Hangzhou Research Institute, which Wang Yuan participated in and led, has successfully incubated many C-end products such as NetEase Cloud Music, NetEase Cloud Classroom, NetEase Yanxuan, Lofter, NetEase Cloud Reading, NetEase Blog, NetEase Photo Album, and many B-end businesses such as NetEase Sufan, NetEase Yidun, NetEase Qingzhou, and NetEase CodeWave. It is the only enterprise research institute that has achieved continuous success in China (similar institutions have the Grand Innovation Institute and Innovation Factory). 数字化转型网（www.szhzxw.cn）
Another core executive of Jiutrace technology is Chen Chengfeng, who serves as the head of engineering research and development of Jiutrace Technology. Chen Chengfeng was the co-founder and VP of Xiaoying Innovation Technology, a company based in Hangzhou. Prior to founding Xiaoying Technology, Chen Chengfeng served as the Technical Director of Hongsoft Mobile BU and the head of MediaPlayer and Camcorder, responsible for the OEM development and delivery of multimedia technology for South Korean mobile phone manufacturers such as Samsung and LG.
Mr Wang decided to leave NetEase about four months ago to start his own business. On May 22 this year, Wang Yuan announced his resignation from NetEase in his public number “Cold technology Hot thinking”. At that time, Wang Yuan said he would return to his familiar to C market. He judged that with the maturity of large models and hardware supply capabilities, generative AI applications will usher in a new window this year and next, especially in the European and American markets.
News of Wang’s venture attracted the attention of hundreds of investors. At present, Jiutrace Technology has completed a seed round of financing, mainly in US dollar funds, and has reserved sufficient funds for the initial team for two to three years. The company is looking for product partners and algorithm engineers. 数字化转型网（www.szhzxw.cn）
The company’s first product is positioned as a “personal data-centric general office assistant” that acts like a secretary to help individuals complete typical knowledge tasks such as attending meetings, responding to messages, answering emails, reading documents, writing reports, and managing schedules.
Wang Yuan judged that the general office assistant market space is huge. For the one billion knowledge workers in the world, according to the proportion of 5-10% paying users, there are tens of millions or even billions of paying users, according to the annual fee standard of 100-200 dollars, there are billions or even billions of dollars in revenue scale, corresponding to the company market value of up to 100 billion dollars. 数字化转型网（www.szhzxw.cn）
However, what the product form of personal office assistant looks like is difficult to define clearly, and many investors are not optimistic about this direction. For example, Zhu Xiaohu, managing partner of Jinsha River Venture Capital, put forward a “violent theory” in an interview in March this year: “Today’s ‘personal assistant’ is the imagination of technical personnel.” I’m just asking you, how many of you need a personal assistant? Typical fake needs!”
Wang Yuan understands this. He thinks many people can’t imagine the shape of this product, which is a good thing, because great products always come from the consensus of a small group of people, and can’t come from the consensus of a majority of people.
There is no long mark technology on the market completely targeted products, the only slightly similar Rewind. However, the product that Jiutrace technology wants to make contains five times the personal data of Rewind – Rewind can only slowly accumulate data by recording the screen, and ten times the functionality of Rewind – Rewind almost only does keyword search function.
Because it involves personal office data processing, it is necessary to ensure that the data runs on the user’s local end-to-side large model, which must rely on the maturity of the AI PC market. Obviously, the AI PC market is still in its early days, and Wang Yuan described today’s AI PC as similar to the iPhone4 just released in 2010, so now is a good time to start a business.
After the first product, Jiuchen Technology plans to launch more Office products with “personal data as the center” as the cornerstone, such as agency management, notes, and so on, and strive to become the “Office in the AI era.”
Like most AI to C products, Jiuchen Technology has carried out a global layout from the first day, and the core is for overseas markets dominated by Europe, the United States and Japan, and takes subscription fees. 数字化转型网（www.szhzxw.cn）
the company’s name “Long Trace” comes from Jobs’ famous saying to make a dent in the universe. Wang Yuan said that Apple, led by Jobs, has created or popularized new product categories in PC, smart phones, tablet computers, online music, TWS headphones and other fields, leaving traces of success. The creation of Long trace technology, Wang Yuan hopes to pay tribute to Jobs and continue the innovative spirit of NetEase Hangzhou Research Institute, leaving more traces in the industry.
Recently, a reporter interviewed Wang Yuan. What does he think about AI entrepreneurship?
First, why start a business?
Q: How long have you thought about starting your own business?
Wang Yuan: In fact, I hope I can do something in 2019, but in NetEase, the various departments I manage the business, it is really quite difficult to hand over, and the boss has been giving me more responsibilities. If I left back then, it wouldn’t be easy for my boss. So it took four or five years.
Q: You announced your business in your public account, saying that you “went to see the world after brushing your clothes”. What does “gone” mean?
Wang Yuan: I am responsible for two major businesses at NetEase. The first one is the back-end such as data center and IT system construction. Last year, the IT system successfully realized the integration of the whole group, and the large-scale self-built data center in Guizhou was also successfully put into operation. 数字化转型网（www.szhzxw.cn）
The second piece is my most core work in recent years, responsible for managing NetEase Sufan’s big data, cloud native, low code, to G and other related businesses in the field of enterprise services. Last year, I basically put the corresponding business department construction and business development to a good state. Compared with starting from scratch a few years ago, it has now achieved business formation, and some businesses have reached the level of first-tier suppliers. So, this part of the business is over.
In general, NetEase Sufan, where I serve as the general manager, has been integrated into the new Digital Intelligence business Division, and other work of Hangzhou Research Institute, which I am responsible for, has also been properly handed over and arranged.
Q: Have you considered doing this within NetEase?
Wang Yuan: I haven’t mentioned it internally. The first reason is that I think the company’s strategy in recent years has been very focused, basically not opening new business lines. Second, NetEase has a large stock of business need me to manage, I certainly do not have enough energy to pull up a product form that has not appeared in the industry, this difficulty requires personal investment is very large energy to do. 数字化转型网（www.szhzxw.cn）
I have learned that many AI startups now imitate American products, such as Gamma fire has a lot of AI PPT products. Imitation is good, of course, but we don’t have a perfect target product, and will go directly to the U.S. market to do zero-to-one innovation. The difficulty and challenge of this job requires my full time commitment.
Q: In the past few years, your core business is NetEase Sufan, which is a business to B, but why do you choose to do C?
Wang Yuan: In fact, in my 18 years in NetEase, the longest time involved is to C business. In addition to NetEase Sufan to B, NetEase Cloud music, NetEase cloud Class and other products are to C. The second consideration is that we want to be a global market, and it will be easier to start to C. In the United States to do to B quite difficult, the formation of offline sales team, management is certainly more difficult than to C. 数字化转型网（www.szhzxw.cn）
Second, why do AI office assistants?
Q: How do you consider the direction of entrepreneurship?
Wang Yuan: We decided to build a C-end application layer product instead of foundation model or large vertical industry model, such as medical treatment. I did a medical large-scale model for nearly a year last year, but I felt that the vertical track was very difficult to develop, and finally it became the customization of the project to B. 数字化转型网（www.szhzxw.cn）
Q: What do you think are some good products for the toC market?
Wang Yuan: There are three main types of products. The first type is emotional companionship and chat, which reflects the emotional value of the large model. The second category, if you look at the a16z Top 50 report, is more in the area of image and video content editing and content generation, 50 applications may account for half; The third category, which appears in the latest a16z report, is productivity products.
Productivity product track has appeared a lot of products, such as the domestic browser plug-in Monica, AiPPT and so on. We belong to the office assistant category of productivity track.
Q: Why is the first product positioned as a general purpose office assistant centered on personal data?
Wang Yuan: The big model has strong general knowledge and intelligent processing power, but ChatGPT’s usage data less than once a day shows that it can’t do much specific work like a novice on the first day of employment, unless your job is mainly to write jargon. Even if the next generation of large models is powerful, it will not be possible to effectively help in most scenarios without understanding the user. Large models will become an increasingly standardized commodity, and user data will become increasingly valuable. In the field of office assistants, those who are good at acquiring, organizing and utilizing user data are in the world. 数字化转型网（www.szhzxw.cn）
Our products are like a secretary who accompanies you all the time for many years, and knows your knowledge, ideas and habits very well, so that they can efficiently help you complete all kinds of typical knowledge work such as attending meetings, replying to messages, replying to emails, reading documents, writing reports, and managing schedules.
Q: Fei Shu and Dingping are also doing similar office assistants.
Wang Yuan: It’s different. Fly book and nail nail are actually specific office software, and the office assistant we are going to do does not do office software itself, but it will use office software like a person. I think the main body of office software is functional, not intelligent. Our focus is to make intelligent products, such as I help you reply to emails, but I will not build an email client; I’ll write notes for you, but I won’t make a note-taking app. 数字化转型网（www.szhzxw.cn）
On the other hand, we mainly target overseas markets. In fact, in overseas markets, the whole family bucket software of Nail nail and fly book is not very popular in the market, and we tend to use multiple more vertical and more professional software to complete the work.
Q: Why did you choose the US market first? Is it related to user payment habits?
Wang Yuan: Generally speaking, the size of the US market and the willingness of paying users are the highest. I am making a terminal software for the TOC market, which only involves personal data, so there is no need to do it in the way of to B, and the business closed loop of to C is simpler and the payment link is shorter. The US market is certainly the largest in the world.
In addition, the American market for software anti-piracy awareness is relatively strong. As a software that runs entirely on a PC, we are technically vulnerable to piracy because it does not rely on cloud services. The American market is generally more concerned about this, especially for office tools, and the average American company does not allow employees to use pirated software.
Our idea is to do it in the US market first, and see how much the domestic market can do when the product is made, but there is really not a high expectation in the country.
Q: Do you have any benchmarked products?
Wang Yuan: There is a slightly similar one called Rewind. Rewind started with a PC that recorded your screen status and the sound your microphone heard in real time. When you think about something you saw two days ago, Rewind will help you find it, and it’s just such a simple function.
My understanding is that screen recording is only a small part of personal data, such as your computer already has a lot of documents, browser has opened a lot of web pages and documents, these data are not captured. In addition, Rewind only does a keyword search function, but it doesn’t do anything else, such as writing emails for you, replying to messages, managing your calendar, etc. So after this product has been on fire for a while, the growth is actually not particularly good. I am also a very early user of Rewind, and the experience feels very low-frequency for me, and I may use it once in a few days. Such a product would not pay very well either. 数字化转型网（www.szhzxw.cn）
Q: It sounds like it’s going to be a big test for data security.
Wang Yuan: This is also a core reason why we believe that we have the conditions to do this at this stage. In theory, I want the product to access all of the user’s office data as much as possible, and I must ensure that the large model processing this data must be in the user’s local end-to-side large model, and absolutely cannot be provided in the way of cloud services.
Therefore, this product must rely on AI PC, and when AI PC appears, our product may appear. The AI PC is still in its early days, a bit like when the iPhone4 was released in 2010. This is a good time to start a business. 数字化转型网（www.szhzxw.cn）
Q: How many investors approve of your product direction? For example, Zhu Xiaohu said, “AI personal assistant is a typical pseudo-demand, which is the imagination of technicians themselves.”
Wang Yuan: I haven’t talked to Zhu Xiaohu. Indeed, some investors’ favorite logic is to ask you which product in the United States is the target of this product, but what we want to do is a completely new product form. If the investor does not have a concrete product in mind, he may not be sure whether this direction is OK or not, and there will be concerns, which I think is understandable. Until a new category comes out, that’s what everybody does. To be honest, when the iPhone came out, there were many people who said that the iPhone was not useful, wasn’t it just calling and texting?
Q: What’s the financing situation now?
Wang Yuan: I have a confidentiality clause in the agreement with the investor. I can only say that all funds are in US dollars, and the entire structure of the company is also oriented to overseas markets, especially the possibility of listing in the United States. However, the company’s production and research team is in the country, and China still has engineer dividends.
本文由数字化转型网（www.szhzxw.cn）转载而成，来源于甲子光年；编辑/翻译：数字化转型网宁檬树。
