After Reviewing Hundreds of AI Hardware Projects, We Don't Think This Is an Era of Parameter Supremacy | Linear View

In the frenzy of AI hardware investment, everyone loves to talk about technology, specs, and ecosystems. But how do you see through to a product's essence and long-term value? The answer may have less to do with those technical parameters than you'd think.

After reviewing hundreds of AI hardware products that have emerged in recent years, the Linear Capital team has arrived at a somewhat counterintuitive conclusion: for many teams, the most urgent gap to fill isn't technical—it's insight into human nature. The industry has grown accustomed to copying winners' features, yet consistently fails to replicate that effortless, "follows your instinct" quality in their products.

Today's share comes from Linear Capital's investment team, reflecting on their recent observations about AI consumer hardware entrepreneurship. They share their views on development trends in AI consumer hardware and their definition of a good product, approaching the topic from the perspectives of "human nature, culture, and scenario"—and proposing a framework of "want to buy, smooth to use, stick around" for evaluation. They also look ahead to the path for Chinese AI hardware to go global. We welcome more friends to exchange ideas with us.

Based on our observations of the AI consumer hardware space this year, we see multiple development paths in this field. Each has the potential to yield unexpectedly successful projects, and each comes with its own challenges.

But one strong impression we have is this: hardware provides the entry point; software determines the ceiling.

Today's share will focus on "how to judge whether something is a good product"—some of our recent thinking on this question.

To get straight to the point: what exactly is a good product? We've distilled a three-layer model, from outside to inside: user scenario → culture → human nature.

The outermost layer is user scenario: the product must hold up in a real moment. Many Chinese consumer hardware products have remained stuck at the level of "stacking features and specs." A truly good product isn't about the function or technology itself—it's about solving a problem or creating a distinctive experience in a real user scenario.

Take, for example, the "Memories" feature in Apple Photos. It plays background music while showing old photos, helping users quickly enter an immersive, nostalgic mood. And when the user presses pause, prepares to speak, or wants to share, the system automatically lowers the background volume—because it "understands" that you're probably about to talk or show something to a friend. These are designs that flow with human nature within real scenarios.

Behind this is the designer's thorough understanding of what the user wants to do in that scenario. So when we evaluate early-stage projects, one of our core judgment dimensions is this: is the founder building a function, or creating a distinctive experience within a real user scenario?

The middle layer is culture: the product must be naturally accepted across different user environments. This is especially critical for teams going global. Apple FaceTime has no beauty filter—this reflects Western cultural values of authenticity and confidence. In Eastern cultures, beauty filters are to some extent a form of social etiquette; people need to confirm the relationship before deciding whether to show their bare face. The same feature can be received completely differently in different cultures. Hardware going global isn't about replication—it's about creating an expression that local users can naturally understand.

The innermost layer is human nature: this is the foundation of the product. We can illustrate this with dopamine and oxytocin: dopamine drives people to constantly seek new stimulation, while oxytocin cements relationships to provide a sense of security in the present moment; a good product that lasts needs to find balance between these two. A good product can't just be addictive—it must foster attachment, sustained use.

At the end of the day, a good product starts from human nature, is nurtured by culture, and delivers in user scenarios. All three revolve around understanding the user, not starting from some technology and "looking for nails with a hammer in hand."

For AI consumer hardware, we've developed a standard evaluation flywheel: want to buy, smooth to use, stick around.

First, want to buy. Ideally, a user should understand "why should I buy this" within 30 seconds. Users don't need to understand specs—they need to grasp at first glance what specific problem it solves.

Take Insta360 GO. In a camera market that's already highly homogenized, it didn't talk about image quality, lenses, or specs. Instead, it redefined the use case—it's a "life recorder light enough to stick anywhere," so users understand immediately upon seeing it: this exists to make shooting easier, more everyday. We believe that good products can, more often than not, make a user fall for them with just one sentence or one video.

Second, smooth to use. A user's first experience determines whether the product gets a second chance. Our standard for "smooth to use": fully operational within 5 minutes, no manual, no learning curve. Open-and-connect earbuds are a classic example—no operation needed, just open and they work. Bambu Lab's 3D printers broke into the mainstream because they were designed so that complete beginners could unbox and start using them immediately, and because they built MakerWorld, a community where users who don't know design can download model files to print.

The problem with many failed hardware products: the initial experience isn't smooth, the process is convoluted, network setup is difficult, prompts are unclear. If a user still doesn't know how to use it after 30 minutes, the product has basically lost any chance of being opened again. The more "lightweight, high-frequency" the AI hardware, the more "smooth to use" must be pushed to the extreme.

Third, stick around. What truly determines long-term product value is often software capability. Hardware provides the entry point, while software determines whether users are willing to use it every day—and this is especially pronounced with AI consumer hardware. For software to make people stay, it must satisfy some underlying need, such as:

• Desire to share: Can the product make users want to actively "show off" their results? For example, Whoop's recovery score calculation lets users naturally share after seeing their feedback—this is software value being "seen" by users.
• Sense of achievement: Duolingo's streak is the classic example: by setting goals and offering various achievement awards, it satisfies users' sense of accomplishment, making them willing to open it naturally at a fixed time every day without thinking.

On quantitative metrics, we pay particular attention to weekly retention, monthly retention, frequency of active sharing, and number of active opens. These metrics together determine whether a product can truly become an AI hardware device that's "used every day."

Finally, regarding Chinese hardware going global, our basic judgment is: we don't necessarily need to copy the Western "soul."

European and American tech giants export a "premium, leading" set of values, emphasizing high premiums and top-tier experiences. Chinese companies can forge another path—we call it "Inclusive Tech Innovation."

The core of this path isn't blindly pursuing premium positioning. It's about letting broader populations access good technology that was once a "luxury" at an acceptable cost. If we can achieve this, what we create won't just be commercial value—it will be a new proposition that can be embraced by more users globally.

Of course, making this path work demands more. It's no longer just about competing on specs or racing to the bottom on cost. It requires us to understand technology, understand product, and truly understand the culture and human nature of different markets. Understanding the target market is prerequisite, but the answer we ultimately arrive at could be completely different from what Silicon Valley companies conclude—and that's precisely how a new path can be forged.

At the end of the day, product and investing alike are continuous practices. Theory can guide direction, but the real path can only be walked step by step in the market. Practice is the sole criterion for testing truth, and it's also the only way to converge on the correct iteration direction for AI hardware.

For more perspectives on AI consumer hardware products and project discussions,

please reach out via email: dunmin@linear.vc