A former DJI employee developed a consumer-grade textile machine and secured hundreds of millions of yuan in financing from HSG, Shunwei, miHoYo and other investors | Product Observation

Author ｜ HUANG Nan

Editor ｜ YUAN Silai

Hu Wenxin, founder of ClawLab Intelligence (CLAWLAB), is a maverick figure in the current hardware track.

With an engineering background, he used to work at leading tech giants such as DJI and Meituan, and has secured investment from top-tier institutions. His resume is solid enough to support him to choose any market-validated popular track, launch products, build business narratives and complete successive rounds of financing in the shortest time possible.

Yet Hu Wenxin dived into a category that was almost ignored by the public in the past three decades: home textile weaving machines.

Few people know exactly what Hu Wenxin is working on. Founded in December 2022, ClawLab Intelligence has deliberately kept a low profile in recent years — it never distributes its business plan to the public, and has not taken the initiative to hold roadshows or conduct corporate promotion. If investors are interested in the product direction, they will be invited directly to the office to see the demo.

Hard Krypton has learned exclusively that in the past few years, ClawLab Intelligence has successively completed several rounds of financing totaling over 100 million yuan. Its Pre-A2 round was led by Vision Capital, with over-allotment from existing shareholder Shunwei Capital. Its Pre-A3 round was led by miHoYo, with substantial over-allotment from existing shareholders Vision Capital and Shunwei Capital. Previous investors include HSG, among others.

The choice of the textile track is based on a very simple insight. "Textile naturally ranks first among the basic needs of 'food, clothing, housing and transportation', which is a sufficiently large and real demand. Customized DIY is a natural extension for users to explore self-expression inward and output personalized content outward after their rigid basic needs are met." In the early stage of starting the business, Hu Wenxin judged that the ceiling of this track is high enough to be worth exploring for a very long time.

The textile DIY track has hundreds of millions of potential users around the world, but there has long been a lack of complete hardware-software integrated solutions. Different from the 3D printing track, household smart weaving machines have no reusable parts and open-source algorithms, and the only reference they can draw on is old-fashioned weaving machines. That means Hu Wenxin has to define a brand new device from scratch.

In 2024, ClawLab Intelligence launched its first product, an automatic tufting gun, which was only used as a verification tool for small-scale overseas markets, and the total revenue in two years was nearly 100 million yuan.

ClawLab Intelligence automatic tufting gun (Photo credit/ Enterprise)

Beyond hardware, what Hu Wenxin has spent a lot of time polishing is a consumer-grade textile Station platform. This is a new-form physical computing terminal that does not require professional pattern making. Users can generate weaving patterns through simple drawing or taking photos, and complete the weaving of various textiles such as scarves, doll clothes, pet accessories, and small plush products.

In Shenzhen, it is impossible to survive in the fiercely competitive market by only making 90-point hardware. The ceiling of startups today actually lies in their software and ecosystem capabilities, and this Station platform has become ClawLab's deepest moat. The company is about to release a new lightweight desktop weaving product, and build a content sharing community to meet the decompression creation and small-scale customization needs of enthusiasts.

Textile is one of the oldest crafts. While the market is still chasing new categories that can be intelligently upgraded, Hu Wenxin has set his sights on a huge blue ocean that has not yet been touched by modern computing technology.

Household Smart Weaving Workstation

Hu Wenxin's decision to anchor on the "weaving" tool category was not out of intuition. As early as the product definition stage, he and the R&D team systematically investigated the four major textile crafts: piercing, hooking, sewing and weaving.

They found that the first three crafts, namely piercing, hooking and sewing, have naturally scattered processes, which can only "add" or "splice" on existing fabrics, and cannot independently produce a complete textile. Weaving, by contrast, can directly use yarn as raw material to form various three-dimensional finished products such as clothes, scarves, accessories and plush dolls. Its highly closed process is more suitable for being integrated into an all-in-one intelligent device, which conforms to the desktop creation logic of "put in raw materials and get finished products".

This difference determines the possibility of application in household scenarios. Although the weaving craft is difficult, it has a complete system that can be deconstructed with intelligent engineering methods.

However, in the past three decades, almost no enterprise has made systematic attempts in this direction. In the 1990s, household weaving machines once had a short period of glory, with established brands such as Singer and Brother occupying the mainstream market share of weaving equipment for a long time, and later many domestic hand-cranked weaving machine products emerged. This kind of equipment relies entirely on pure mechanical structures, and all operations need to be done manually.

The seemingly simple weaving action is actually a dynamic process highly dependent on timing — the linkage of different pattern colors, different knitting needles, and lines of textures. Any deviation in tension or movement will lead to the scrapping of the entire textile.

Therefore, to build an automatic weaving device, the primary problem to be solved is not material stacking or fixed-point processing, but how to build a dynamic control system for real-time perception, adjustment and compensation.

"Many people think that weaving is nothing more than turning thread into cloth. It was not until we started working on it that we found every link is an independent technical problem." Hu Wenxin told Hard Krypton, "The mechanical action combinations of plain stitch and cable stitch are completely different, the tension control curves of thick wool and thin cotton thread are completely different, and the contour forming logic of a piece of clothing and a hat is completely different. These differences cannot be solved by debugging, and must be clearly defined at the beginning of system design."

The first hurdle facing ClawLab Intelligence is the dual lack of technology and data. There is no open-source algorithm for reference, no ready-made data set for training, and no mature supply chain for reuse. Even the most basic question of "how the weaving action should be digitally defined" has never been answered before.

Hu Wenxin and his team spent nearly three years breaking down the weaving craft into programmable control algorithms step by step, forming their own technical barriers.

The team's solution is to migrate AI hardware capabilities such as robot control, motion planning and computer graphics to product R&D. After one and a half years of work, they finally ran through the first version of the reliable weaving process.

ClawLab Intelligence programmable control algorithm (Photo credit/ Enterprise)

But for a device to enter households from a "tool-type" product, the real challenge lies in the user interaction layer. They must lower the design threshold to a sufficiently low level, so that people with no weaving experience at all can get started, otherwise the product can only stay within the professional circle.

The pattern making of traditional weaving relies heavily on experience. A qualified pattern maker needs 5 to 10 years of training to understand patterns, master the stitch logic and adjust yarn parameters.

The ClawLab Intelligence team is also independently developing a textile AI Agent, which is based on a domain knowledge-enhanced vertical model, with built-in pattern making algorithms and compilation capabilities accumulated by the team. After entering the weaving scenario, users do not need to understand complex stitches and pattern parameters. After uploading images, a design Agent that can understand what users want can automatically analyze the textile style, extract features, and generate a design scheme that can be directly sent to the machine for weaving.

Users only need to describe their design intention in natural language, and the AI Agent will complete all subsequent weaving work. The work that traditional pattern makers need several days to complete is compressed to a few minutes in this all-in-one textile Station, forming a complete closed loop from "user intention" to "physical finished product".

This is perhaps the most interesting part of this category. On the one hand, it is a blue ocean market that has not been verified yet but has huge potential. On the other hand, it has long lacked systematic intervention of modern technology, leaving a huge technical gap. But precisely because it is a niche field, once the full-stack system is built, the width of this moat far exceeds what the single-product hardware logic can understand.

The deeper attraction comes from the track itself. In Hu Wenxin's view, this is not a story of a single category, but has the potential to become a platform. "Only for the weaving machine device, based on its underlying technology base, we can derive a variety of differentiated product definitions for different segmented scenarios. We believe there are still a large number of untapped scenarios and categories in the textile track waiting to be explored."

An underwater market with over 100 million users

Around the world, the popularity of weaving has continued to rise in markets such as Europe, America and East Asia, with the core active enthusiast group reaching the scale of tens of millions, and the broad potential coverage of users can exceed 100 million.

Public social media data shows that on Xiaohongshu, the total views of topics related to "weaving girls" are nearly 900 million, and the total traffic of related topics such as crochet weaving and hand weaving exceeds 3 billion. On TikTok, the views of the "crochet" hashtag have reached more than 200 billion times.

A sufficiently large user base is only the first step. More importantly, there is no need for market education for the end products of weaving. The warmth of scarves, the companionship of hats, the softness of sweaters, and the healing power of plush toys, people's yearning for such emotional products is natural and self-evident.

Weaving products produced by ClawLab Intelligence's weaving workstation (Photo credit/ Enterprise)

"The weaving works displayed by users on the platform are just the tip of the iceberg. The voices in the comment area that say 'it looks good', 'I want it too', 'is there a tutorial' are the real underwater demand," Hu Wenxin said. "We are not creating a new demand. At present, no product can meet the demand of these people for a custom sweater or a pillow with a specific pattern. This is exactly our entry point."

The breakthrough point is to leave the repetitive labor, pattern making and weaving processes that take weeks to the machine, so that users can retain the emotional value of designing, choosing and creating by hand, greatly reducing the time and threshold from wanting a product to owning it.

Hu Wenxin told Hard Krypton that this compensation method can give rise to a unique user evolution path. A large number of textile DIY users initially enter this field driven by emotional factors: decompression, killing time, and enjoying the satisfaction of making something by hand. With the accumulation of skills and the influence of the community, some people begin to extend from hobbies to the light-weight commercial stage, taking small-batch orders.

Around this evolution path from emotional consumption to light-weight business, ClawLab Intelligence divides user portraits into several penetration directions.

The first batch of seed users are heavy users of existing weaving equipment. They are either enthusiasts of old-fashioned weaving machines for many years, or individual creators who get stable income from hand weaving. They have a cognitive basis for product tools, and their pain points are direct and specific: low efficiency of manual pattern arrangement, difficulty in realizing complex patterns, and unstable yield. This type of user does not need to be educated about "what weaving machines can do", but needs products that can truly solve daily problems.

ClawLab Intelligence automatic tufting gun (Photo credit/ Enterprise)

The second type of target users are small B-end operators, such as doll clothes makers, pet clothes weavers, and people taking holiday customization orders. Their core demands focus on efficiency and customization capabilities, and one device can effectively shorten the creation cycle of a single product. The value of this group is not only limited to purchasing equipment, and the finished products they produce will naturally attract the attention of C-end users, forming penetration from the B-end to the C-end.

Ultimately, ClawLab Intelligence's goal is to cover ordinary mass consumers, who have no DIY creation needs and do not care about the weaving process, but only care about whether they can quickly get the textiles they want. At this stage, its products will break away from the positioning of creation tools, and truly land as a small flexible supply chain in household scenarios, realizing instant customized delivery of textiles.

"This group of people is essentially in a process of continuous upward movement of demand levels. In the first stage, users' pain points are solved. In the second stage, efficiency is improved. In the third stage, the demand for good-looking products is met. In the fourth stage, users can get whatever they want," Hu Wenxin analyzed.

In the past three decades, household weaving, which used to be a productivity demand of "one sewing machine per household", has gradually faded out of people's daily life. Spending weeks hooking a blanket has been abandoned by most people in the fast-paced life. The industrial age brought weaving into factory assembly lines, and standardized scarves and sweaters flowed to every shelf at the lowest cost. The efficiency has been greatly improved, but the cost has also followed: the individual attributes originally carried by weaving have been completely compressed. The specific person, specific aesthetic choice and specific expression intention behind each textile have all been simplified to "the same style".

But the emotional need to create by hand has never disappeared. With the maturity of the hardware supply chain, the decline in the cost of motors and sensors, and the transferability of AI capabilities, it is for the first time feasible to "turn a complex weaving device into a consumer-grade product".

"Everyone has something they want to express. A favorite pattern, a preferred color, a unique touch that belongs to them, and the memories and emotions attached to it. These ideas should not be blocked by 'mass production'. So ClawLab Intelligence has made a weaving machine that can be placed on the desk. It is not a toy, not a concept, but a machine that can truly weave finished products," Hu Wenxin said.

Users' designs grow out of the screen, turning into touchable textures, the warmth that can be wrapped around the shoulders, and unique gifts that can be given to others. The problem solved by this machine is not to make things faster or cheaper, but to make textiles belong to specific people again.