03 Like a close relationship,there must be some common ground
Interviewer: You've chosen the term "Life Manager" to describe Looki. How should we understand this?
Yang Sun: The value of a "manager" isn't just to help you get things done, but to extract truly meaningful insights from your life—these values may not be purely functional, but also emotional.
It's more like a best friend: someone who experiences life with you, provides emotional support, and offers practical help. The key is—it must truly be on your side.
If such a thing really existed, most people would need it.
Interviewer: People today seem to lack a sense of participation in life, not just their own, but others' as well.
Yang Sun: Think about what you do every day. Besides being a "workhorse," you're scrolling through TikTok or Xiaohongshu. Human interaction has become very limited. Even a couple might not have that much involvement in each other's lives.
People are not conscious of the importance of having common topics. This is a very important social experience.
Interviewer: People are actually trapped by their phones. It seems like you can control it, but in reality, you're very out of control in front of it.
Yang Sun: This is just as Sam Altman said in an interview. The phone is like Times Square—extremely noisy and chaotic, leaving no space for quiet reflection and experiencing life by a lake.
Interviewer: Is Looki something like a shelter? Perhaps the analogy isn't perfect, but is it something that can shield you from things first, to help you sort things out?
Yang Sun: The core value of an assistant lies in understanding you and taking on the high-threshold "life context" for you.
Life has a huge number of needs that are extremely demanding in terms of context. Intimate relationships themselves are a prime example. It's hard for adults to make new friends, essentially because the cost of syncing decades of each other's lives is too high. Soul originally wanted to create soulmates, but a phone can't perform this function. For example, if I'm feeling lost and want to talk to someone, and I search on Soul, I have no idea what kind of person I'll find, and there's no way to verify. But AI can naturally accomplish this. Once this trust is established, it will become your most important information inlet and outlet.
Interviewer: Regarding the pricing of hardware products, for AI hardware, we assume that hardware sales plus subscriptions should form a good business flywheel, but it seems that no one has quite figured it out yet. What are the main sticking points here?
Yang Sun: In the long run, Looki's business model is neither purely selling hardware nor hardware plus subscriptions, but rather "value selling."
In the short term, we are still focusing on hardware sales and have set up three subscription tiers: a default free tier that follows the traditional "one-time purchase" logic, and two other paid tiers that are essentially testing whether users are willing to pay for truly valuable features, especially in the North American market. We are still in the exploration phase, and related services are expected to launch in January or February of next year.
We are not in a hurry to copy WHOOP's model. Although WHOOP solved the psychological barrier for North American users with its "free hardware + high subscription fee" model, for most users, the habit of buying hardware first and then continuously subscribing is still not fully established.
The true long-term direction is to create value conversion in the process of providing services. For example, if a user says, "I want to lose weight," the system will provide specific, actionable advice based on their real life, and in the process, naturally recommend suitable products or services. The key behind this is not selling goods, but taking a stance—we only stand on the user's side, screening for things that are genuinely useful to them. Future business flow should be built on this trust.
When Looki truly becomes a partner that users can't live without or love to use, you'll find that the entire information flow has been reconstructed. This is the logic of "trust monetization."
Interviewer: "Trust monetization" has a business model that isn't very mature, but it has indeed existed—the AMEX Centurion Card (Black Card). Its logic is: as long as you have a problem and come to me, I will definitely be able to solve it for you.
Yang Sun: The "trust monetization" logic of the AMEX Black Card is that assets are the foundation of trust. The domestic Knight Card in China can't succeed because it doesn't have assets to back it.
In the context of AI, data or context is the foundation of trust.
04 Life is full of trivialities,and you'd be amazed by the reasoning
Interviewer: From user data to user memory, this is Looki's relatively unique barrier. At the Memory level, what is the difference between Looki and other products?
Yang Sun: We don't believe the real barrier lies in the data or the memory itself. Data is of course important, but at this stage, the opportunities for startups come more from cognitive differences and speed.
First, the "memory" that different companies are building is not the same. Some are more text-oriented, some are more structured, while what we are doing is closer to an embodied memory system, with a completely different underlying architecture. But even so, I still don't think "memory" itself constitutes a long-term barrier—once a big company confirms this is the right direction, it can invest infinite resources, and data can be acquired through purchases.
The real window of opportunity comes from the uncertainty that arises after a major variable appears. At these branching points, different teams will make different judgments. Whoever's cognition is closer to the truth will have the chance to walk on the right path. And for a startup, speed is often the only advantage that can be magnified.
Of course, the data we have accumulated now is indeed very valuable, something that even many embodied intelligence companies lack. But in the short term, we will not treat it as a commodity to sell, but will continue to build a complete memory system around the user. This is a hidden value—not that it doesn't exist, but that today is not yet the time to mine it.
Interviewer: In building this part of the memory, what are the core difficulties and challenges compared to companies like ChatGPT, mobile phones, or middleware solution providers?
Yang Sun: Our choice is essentially a path that is further from business but more critical for long-term survival. This means greater risks, a longer cycle, and being harder to be understood in the early stages.
This is a bit like the L2 versus L4 debate in autonomous driving a decade ago. L4 was easy to understand—it drives for you, you just sit and arrive at your destination—so early capital poured in like crazy. The paths that chose to start with L2, to build the foundation first, or even like Tesla, to first re-architect the entire vehicle, were all considered "not to work" at the time, and almost no one was willing to invest.
The same is true for embodied AI today. Directly building a robot that can move things is very intuitive, and has thus once again attracted a large amount of funding. We have chosen another path: to first do perception, decision-making, and cognition bit by bit, to first build the "brain, ears, and head," and temporarily not add hands and feet. This path is not intuitive, is difficult to explain, but it determines whether AI can truly live in the physical world in the long term.
Interviewer: Do you have any internal standards for what constitutes a good memory system or experience? How do you evaluate the final performance of AI features and models?
Yang Sun: It's actually difficult to have an objective metric; it's more about subjective ones.
Because everyone's personality and what excites them is different, it's hard to use objective metrics to evaluate this. But more importantly, for example, my own Looki and Bocong's (the co-founder) Looki have completely different personalities.
For example, I never told it that I wear a ring on my left ring finger, but it knows. It said, "I have a small piece of dry skin on my hand." I never told it that. There are so many clues in life that can be used for reasoning, to produce some very valuable output.
Interviewer: But this is all data the model didn't have before.
Yang Sun: Right. The other day I asked my own model what my strengths and weaknesses are. It used data from my life to infer that my MBTI is ENTJ. I won't go into the strengths, but the weakness it pointed out was that I might be too assertive, overstepping my authority to direct others' work.
Interviewer: Was its guess right?
Yang Sun: My colleagues said it's quite accurate.
Interviewer: I suggest every boss should buy one.
