李一豪
Key Views & Dialogues
After OpenClaw, Who Will Define the New Battleground for Proactive AI? | A Conversation with 黄柏特 of AirJelly
AirJelly’s differentiated bet is to use Enter to capture high-value intent, turning noisy screen recording into task- and event-based long-term memory. OpenClaw-related execution makes that memory actionable across applications, but retrieval quality, privacy trade-offs and a one-to-two-year window before giants enter remain the key risks.
View summary (7)
AirJelly’s core bet is not capturing more screens, but using Enter to capture the highest-value user intent. My Context took screenshots every 15-30 seconds and analyzed them every 15 minutes; AirJelly samples the moments when users press Enter in IM conversations, Chatbot dialogues and browser searches, then turns a “chronological history” into a “biographical record” organized around tasks and intent. 黄柏特 argues that the scarce input for proactive AI is not full-session recording, but “capturing the highlights.”
The real product leap comes from combining Context Memory with Agent execution, not from recording or automation alone. After integrating with OpenClaw-related frameworks, AirJelly can invoke skills, operate computers and browsers, and continue pursuing leads using cross-application memory; it can submit PRs to improve itself and once used a memory of a Boss Zhipin interaction to trace back to a local WeChat image and recover a résumé that ordinary file search could not find. This is what 黄柏特 calls “1+1 greater than 10.”
The team found its moat during a product pivot after its platform swallowed the original direction: Claude Code could quickly cover the execution layer, but was hard-pressed to replicate cross-application Context. In December 2025, the team bet on task engineering; Cowork and Claude Code’s subsequent shift toward tasks validated the thesis while rendering much of the prior work nearly useless. The resulting startup filter is blunt: if vibe coding can already produce a 60-80 point product, it is not worth pursuing; AirJelly’s Context approach could reach only about 30 points, “which is exactly right, because that is what gives it a moat.”
AirJelly’s contrarian view of proactive AI is that it should converge rather than diverge: extend the user’s current task instead of constantly generating new information. True proactive behavior requires both a clear intent and the relevant context, then uses task progress to infer the next step; prompt timing should reflect shifts in attention, such as switching applications, while signals like dismiss and got it calibrate frequency and keep “proactive” from becoming a cognitive burden.
黄柏特 sees first-mover memory, engineering detail and user acceptance of privacy trade-offs as the moat, but acknowledges that the window may be only one to two years. Private memory accumulated after one month or three months of use is difficult to migrate, while screenshot understanding, event merging and retrieval still produce plenty of bad cases. The bigger risk is moving too slowly: if giants enter before memory has accumulated, “not fast enough” becomes the primary failure path. The balance between privacy and efficiency is, in his words, “an art of getting the heat right.”
The general-purpose route requires a team capable of defining a new framework, controlling resources and marketing globally. 黄柏特 wants to follow Manus’s model: bring users in first, then let real-world use narrow the product toward concrete scenarios, an approach he describes as “respecting users’ wild ideas, then respecting AI and trusting what AI can do.” 李一豪 cautions that most founders may be better served by entering through a vertical, using a new framework to replicate the expertise of a small number of specialists—in essence, “rebuilding a person.”
This is an early-stage bet with high execution risk and strong first-mover effects: a 24-year-old founder, an 8-person team, an angel round recently completed and a second round underway. The end state is not another tool, but a long-term network of “one Agent per person”—Agents that represent an individual’s skills in production while also offering companionship because they hold a complete memory of that person.
In a “Slow, Expensive, and Fragmented” Market, AI’s Opportunity Is Most Seductive | A Conversation with Mizzen AI’s 孙克强 & 李一豪
Mizzen targets user research as the bottleneck left behind by faster AI development, claiming 100x speed and one-tenth the cost in an $80B–$90B fragmented traditional market. Its moat is not replacing interviewers, but training a questioning model on real interviews; continuous small experiments could expand demand, while respondent simulation, brand trust, and research quality remain unresolved.
View summary (7)
- Mizzen AI is not betting on a “cheaper interview tool”; it is targeting the key link in the AI product loop that has yet to accelerate. 孙克强’s own words: “AI has dramatically accelerated product iteration, but user research is lagging far behind R&D.” A traditional project takes roughly 60 person-days and costs about 100,000; Mizzen Insight claims to deliver 100x the speed at one-tenth the cost, compressing the research cycle from months to hours—ask in the morning, get the report after lunch. The team currently has just 8 people. Traditional user research is labor- and knowledge-intensive; adding headcount increases management overhead, while ROI declines when delivery quality stays unchanged.
- 李一豪 sees a large aggregate market with no dominant player, entrenched structures, and extreme fragmentation—a clear window for substitution. He estimates the global traditional research market at roughly $80B-$90B and emerging UX tools at $3B-$5B, while even the leading companies generate only $2B-$3B in annual revenue. His comparison is the $400B traditional legal market and $60B-$70B legal-tech market: fragmentation does not prevent the No. 1 player from going public and generating substantial returns. More importantly, these traditional, semi-digital industries change slowly, giving new entrants room to build an AI-native advantage.
- The real product moat is not replacing interviewers with ChatGPT, but training a questioning model that can get to the crux. 孙克强 concedes that general-purpose models are “very good answerers, not good questioners,” while most high-quality interview processes remain closed-source. The team is securing exclusive interview data that has passed its confidentiality period and using reinforcement learning to build its own environment and reward model, evaluating moderator questions through dimensions including answer authenticity, detail, subjective analysis, and consistency. 李一豪 sees this as the core moat of vertical AI: domain benchmarks and an edge-data flywheel, rather than simply calling a stronger base model.
- The biggest value of lower costs may not be taking existing large projects, but turning user research into small experiments that can be launched every day. 孙克强 observes that product-building costs have “fallen dramatically, even below distribution costs,” pushing customers from one large report a month toward “continuous, incremental user research.” Consumer-electronics customers are already testing small questions such as whether to adopt “try before you pay” and whether handwriting and screen typing should be separated. The platform charges by project with no platform fee, with the goal of “democratizing user research”; Koji believes new-market opportunities are more startup-friendly.
- The counterintuitive advantage of an AI moderator is reducing interpersonal performance, but response quality still depends on mechanism design. 孙克强 acknowledges that respondents initially feel uncomfortable, but argues that AI reduces interference from a moderator’s gender, emotions, and chemistry. The platform discloses its scoring criteria—authenticity, detail, subjective analysis, and consistency—in advance and links scores to incentives, making “more truthful, more willing expression” the optimal strategy. Its respondent supply comes from 2 domestic and 2 overseas vendors plus partner expert pools, with total capacity claimed to be in the tens of millions.
- The team explicitly rejects the shortcut of “interviewing AI” for now, while preserving a long-term path toward respondent simulation once enough real data has been accumulated. 孙克强 says the gap between current models and real human context remains “huge”; fixed training contexts can reproduce only past and incomplete information, while general reasoning and brand trust are still insufficient. The long-term plan is not to generate personas from scratch, but to distill a respondent’s past statements into a preference graph and run evidence-based, low-cost experimental simulations.
- This early-stage bet is simultaneously on a long-term technical thesis and the founder’s ability to pivot. 孙克强 moved from visual preference research and HPS through GPT-3.5 and RLHF to multimodal human preference, then abandoned the art-market direction after studying galleries, artists, and traders. What 李一豪 values is that “real sampling,” rapid pivots, and resilience are strengthened rather than crushed by negative feedback. The endgame is for Mizzen to become “Iron Man armor for every PM,” with Web coding and web user research as its two hands: humans contribute intuition, aesthetics, and “strokes of genius,” while AI completes the research-to-iteration loop. Asked how he would invest $3M across 3 people, 孙克强 said the first 2 checks would go to Quick Stone’s 2 longtime shareholders, and the third to Silicon Valley company Krea.
Your Next Wearable: A Health Companion Around Your Neck? | In Conversation with 潘宇扬, Founder of AI Necklace Odyss, and 李一豪, Partner at CreekStone
潘宇扬 sees glasses constrained by comfort, battery, aesthetics and prescription costs, positioning Odyss as a neck-worn multimodal device that continuously captures food and exercise. A low-frame-rate AI camera, edge-cloud processing and privacy limits support 18–20 hours of battery life, but category demand, branding, regulation and user conversion remain key risks.
View summary (7)
- 潘宇扬 is betting not on the novelty of an “AI necklace,” but on the idea that once glasses hit limits on comfort, battery life, appearance and prescription-lens costs, a neck-worn device may be the more first-principles multimodal entry point. In his view, Ray-Ban Meta’s several-million-unit annual sales rely mainly on the overseas sunglasses market and Ray-Ban’s distribution and brand, rather than proving that AI glasses have become a mass-market terminal; a necklace sits at the front of the body, can see and hear, and places roughly 50 grams on the part of the body best suited to carry weight. “Products should serve people,” rather than forcing users to wear glasses so AI can “hear what I hear and see what I see.”
- Odyss’s core function is to automatically record “every bite” of food and exercise, turning the full path from energy entering the body to being expended into data that can be planned around. It does not seek human-viewing image quality, but uses a low-frame-rate, low-resolution, high-contrast “AI camera”; the team found that 10 FPS delivered no recognition-benchmark gain over 30 FPS, then combines T0-to-T1 load prediction, on-device compression, small cloud models and asynchronous queues to trade limited bandwidth for a target battery life of 18–20 hours. “Every bite you eat is on the record”; even drinking instant-noodle broth can be inferred from changes in the liquid level and IMU motion.
- The team ruled out a general-purpose recorder from day one because hardware needs to give consumers a clear reason to buy, while context can only be a byproduct of a good experience. 潘宇扬 believes general AI is often just a “general-purpose illusion” assembled from multiple optimized vertical agents; meeting notes ultimately belong to the phone OS, and an ordinary person’s life may not be worth archiving all day. More importantly, many products summarize personal information into a system prompt without being able to verify, user by user, whether the agent actually improved or worsened, so context will ultimately prove itself in verticals such as health and coding.
- Odyss’s overseas entry point is not simply “weight loss,” but a mix of biohackers, GLP-1 users and people already at risk of chronic disease. 潘宇扬 cites market observations that roughly 190 million Americans have at least one chronic condition, while the global smart-wearables market has nearly 1 billion users; GLP-1 users eat less but may favor high-fat, fried or roasted foods when choosing a small amount of food, while people with chronic-disease risks need to identify hidden ingredients such as gluten, lactose and purines. The US defines obesity at a BMI of 30 versus 28 in China, yet some US interviewees with BMIs of 30 or even 40 still considered themselves “very strong,” so the product is selling health status and work efficiency, not appearance anxiety.
- Compared with photo-based calorie-counting apps, the necklace’s incremental value is not guessing the dish name one more time, but removing the friction of frequent logging and recovering the time information that a single photo discards. A single image cannot reveal a bowl’s depth, food underneath or how much of a shared dish someone actually ate; it also cannot show whether vegetables or carbohydrates came first, or how quickly someone ate. The team’s consensus is that ChatGPT may currently be the most accurate recognition app, but the limits are still inherent in the input. When the host wore a continuous glucose monitor, a bowl of beef noodles delayed until 3 p.m. drove blood glucose to 17–18 and left him asleep for an hour, showing that the valuable object is the behavioral chain, not just the calorie total.
- The criticism that the product is “anti-human” instead pushed the team to remove real-time interventions, retain only data presentation and future planning, and reduce privacy risk through product boundaries. 潘宇扬 says the harshest criticism came mainly from investors rather than users; his response is that humanity’s descent from the trees was itself “anti-ape,” and that being anti-human only limits the ceiling of the user base—it does not eliminate commercial value. Odyss has no gallery, users cannot view complete video or audio, and see only a health summary; most of the time, the underlying data is discarded after the model produces the summary. The hardware will also include an easy way to disconnect. The team will not interrupt users halfway through a meal to say, “You’ve eaten too much.”
- Consumer health hardware also succeeds or fails on whether it becomes an identity symbol, which is why Odyss wants to pursue a brand position that is “more Oura than Oura.” The name draws on the Odyssey-like idea of adventure and self-transcendence; jewelry qualities, product quality and social currency take priority over medical seriousness or Duolingo-style pure gamification. The team has produced nearly 10 industrial-design concepts and used scoring by US and UK target users, landing pages and ad tests to narrow them down. The company now has roughly 10 people, has completed a first financing round worth several million dollars and is in discussions for a second; it plans to launch overseas as early as Q2 next year and no later than Q3. 潘宇扬 believes the core technology can be validated before launch; the biggest uncertainties remain demand for a new category, geopolitics and regulation.