In 2025, the AI hardware track is buzzing with activity. From AI glasses in various forms to OpenAI's rumored screenless device, countless players are trying to break free from the shackles of the smartphone. The underlying logic is clear: today's AI doesn't lack intelligence; it lacks a "sense of presence."
Trapped in a dialog box, a large model is like a "brain in a vat." When you need to review a meeting or analyze a child's emotions, you're forced to compress vivid scenes into dry text descriptions. This reliance on "manual prompts" is not only inefficient but also counter-intuitive. We are forced to act as AI's "human sensors," while the large model, lacking Context, remains in a state of extreme hunger for physical data.
"In the physical world, humans are essentially passive creatures. We want to make AI proactive," says Looki founder and CEO Sun Yang, hitting the nail on the head.
It is with this thinking that a team with autonomous driving DNA in its veins is attempting to reapply the "perception to decision" logic used for analyzing road environments to everyday life, filling the cognitive "blind spot" that large models have regarding the physical world. The Looki L1 they created, after building its initial reputation in the North American market, is officially launching in China today.
This AI-native multimodal wearable device aims to convert real-time audio-visual signals into model context, allowing the physical world itself to become the most efficient prompt for driving AI's thoughts and services for you.
I. Beyond RAG, Building a User-Exclusive "Data Flywheel"
From "multi-dimensional data" to "structured memory," the real barrier lies in constructing a user-exclusive Context Engineering.
The Looki team applies the "data flywheel" logic from autonomous driving to reshape the user's personal memory: from collecting data at the hardware end, to establishing multi-level data indexing and multi-task decomposition, and finally creating an efficient Agentic System that allows AI to call upon your life data with indexes and priorities, just like a human brain.
According to CTO Liu Bocong, the originality of this system lies in challenging two major pain points of current multimodal models:
First is "long-sequence data understanding." To address the common problem of large models "hallucinating" when processing massive streams of multimodal data (e.g., in the hundreds of thousands), the team, through engineering optimization, enables the AI to accurately understand and connect time slices spanning very long periods.
Second is dealing with "multimodal data Context explosion." Querying video and images consumes a huge context window. Looki has built an efficient context management mechanism in the cloud that can accurately extract key features from massive amounts of data based on query needs, rather than crudely feeding all the material into the model.
Of course, the foundation of all this is data privacy and security. Looki ensures the security of data flow through on-device privacy filtering and a manual upload mechanism via the app. Ultimately, slices of the physical world are transformed into a structured "cloud-based memory bank."
If GPT is the context for your work, loaded with humanity's general knowledge, then what Looki is building is the private context of your life, loaded with your unique experiences and trajectory.
II. From "Manual Prompt" to "Proactive AI"
When the context is rich enough, interaction will eventually evolve into proactive "prediction."
The current consensus in Silicon Valley is that we will eventually see a paradigm shift from Chatbots to Agentic AI. Industry forecasts show that the autonomous AI agent market will experience explosive growth in the next three years, with Memory and Context being the core moats of this evolution.
Looki's ambition is based on these two elements, using wearable hardware to build an entry point to "proactive intelligence."
Currently, the most visible feature Looki delivers to users is "AI automatic editing." It precisely targets the pain point of users who love to record but are deterred by post-production editing. While the user is wearing and shooting, Looki's algorithm automatically identifies the highlight moments of the day and generates a vlog with a narrative logic.
Based on this layer of understanding, the real qualitative change lies in "implicit insights."
As the memory bank accumulates, Looki will no longer be a recorder passively waiting for commands, but an Agent capable of reasoning. It will not only record your diet but also proactively issue a "calorie warning" based on your all-day intake before you order a late-night snack; or provide rational communication advice based on the conversation context when you are emotionally agitated.
This interaction, based on real-time physical feedback, allows AI to evolve from "passive Q&A" to "proactive resonance." This is the core value that distinguishes AI hardware from mobile apps and enables it to truly take over a user's life.
III. Epilogue: When the Physical World Becomes the Prompt
In the vision of the Looki team, the hardware form was never the end goal, but rather a "data interface that grows in the physical world." This reveals the essence of Looki: whether it is a pendant today or evolves into glasses or headphones in the future, its core mission has never changed—to solve the physical world's data hunger and help users accumulate their most valuable private data assets.
In the AI era, computing power will be diluted by Moore's Law, and models will be iterated by stronger versions, but the traces of your life in this world, the scenery you've seen, the conversations with your family—this Personal Context constitutes the unique you. And Looki is what helps you transform this asset into a "digital intuition" that AI can understand and empathize with.
When AI has a "sense of presence," it is no longer a cold tool in the cloud, but a "second brain" in your life.
The most powerful Prompt is, in fact, your unspoken life itself.
