Summary

StressWatch reached nearly 4M users worldwide in two years, was featured by the App Store in roughly 170 countries and regions, and supports a nearly 10-person team that is largely self-sustaining. Growth came from both platform promotion and the product’s inherent shareability: one user’s date ending with her Apple Watch “signing up” for her shot to No. 1 on Xiaohongshu’s trending list; a discussion on Taiwan’s Threads pushed the app to No. 2 overall, ahead of ChatGPT at No. 3.
The product sits at the intersection of 3 opportunities: a growing Apple Watch installed base, HRV already gaining traction in the US but still little understood in China, and incumbent stress products that mostly report a cold number and stop there. Patrick saw “stress” as a distinctly East Asian concern, so instead of repeating “your stress is 85 today, that’s it,” StressWatch uses weather, characters and animated expressions to show the body’s state—turning a roughly 60-point, male-oriented tool into a more emotional, female-leaning 80–90-point product.
Patrick’s core entrepreneurial conviction is that small teams should not fear small demand; the best opportunity is serving the 80% strongest needs of 20% of the population. WHOOP targets obsessive athletes, charging roughly $200–300 for the hardware and about $20 a month; Oura likewise combines hardware with a roughly $6 monthly subscription. A small team can do very well with just tens of thousands of users willing to pay consistently, while a large company often has no interest in a market too small to support its growth targets.
StressWatch’s biggest detour was not failing to find PMF, but underestimating it after finding it—expecting only 100K–200K downloads and diverting roughly six months of effort to other projects. With daily revenue initially in the tens of dollars and the team constrained by the Apple Watch, they assumed the app could only cover a living; during that six-month window, competitors copied the product or added stress features and at one point grew faster. Patrick’s postmortem is blunt: “If a product has clear PMF, don’t hesitate or waver.”
Rather than build a product first and search for a market later, Patrick advises small teams to find the market first and then build the product—in other words, look for “MPF.” He scans category rankings from No. 200 to No. 500: the range proves demand exists but usually lacks a dominant player; if people still use a product with a terrible experience, there may be a real opportunity. Being just 5–10 points better than a mature competitor rarely offsets switching costs; whether early group chats sustain organic discussion, whether creators will promote it for free, and retention benchmarks of roughly 30%–40% on day 1 and 10%–15% on day 7 are more meaningful signals.
The growth window matters as much as the product window, and utility apps are fundamentally cash-flow businesses that require sustained user acquisition spending. StressWatch was roughly a year late in scaling overseas ad spend, by which point acquisition costs in many markets had risen 2–3x; during the first major push, Patrick slept badly for nearly a month until he confirmed that the money spent would at least come back after two months. As scale grows, CPI can rise and subscribers can churn, with the ultimate ceiling determined by whether new users can continue to cover lost users.
The structural opportunity today is not simply to “build an AI product,” but to combine global infrastructure, ad platforms and distribution channels with AI-powered translation, user communication and development iteration to materially lower the cost of global experimentation. Patrick does not see foundation models as a fit for himself, and notes that startups lack user data at launch; AI’s more important role is increasing the throughput with which an organization can test bad ideas. “100 bad ideas” can thus become “1,000 bad ideas, 1M bad ideas”—with the real organizational advantage still residing in people and the mechanics of experimentation.

Deep dive

1. StressWatch Turned a “Niche Watch Utility” into Nearly 4M Users

Patrick defines StressWatch as an Apple Watch-based product for stress monitoring, health management and watch-face decoration. Roughly two years after launch, it had nearly 4M users worldwide and had been featured by the App Store in roughly 170 countries and regions.
Its most viral moments were not medical explainers but everyday emotions. One user posted that her date had ended with her Apple Watch “signing up” for her shot, showing how her stress spiked during the date and StressWatch alerted her; users flooded the comments asking for the app’s name and posting their own dating stress, eventually pushing the topic to No. 1 on Xiaohongshu’s trending list.
A discussion on Taiwan’s Threads pushed the product to No. 2 overall in the App Store, with ChatGPT at No. 3. Japan, South Korea and Germany also began showing solid traction, convincing the team that Apple Watch utilities were not merely a small domestic business.

2. Patrick’s Career Kept Expanding the Boundary of Who Gets to Set the Question

During university, Patrick joined a tiny advertising agency focused exclusively on the fishing industry. Its clients had advertising needs but could not afford large agencies, showing him early that niche industries could still leave room for a small team to survive.
At Meizu, he mainly handled the visual and interaction design of the phone OS. To understand how needs emerge and how products are built from zero to one, he moved to Frog Design, following the design-thinking process through user research, insight generation, ideation, prototyping and testing. He also worked on a car-washing robot roughly the size of a refrigerator.
Agency work could produce satisfying work, but the brief always came from the client. Patrick eventually grew tired of it: “I don’t want to be the person solving the problem. I want to be the person asking myself what problem to solve.” That became the direct reason he left Frog and moved toward startups.
At smart-fitness company Fiture, he worked on a fitness mirror combining a reflective display, cameras and AI motion recognition, and spent a year in New York to support the US business. The experience gave him exposure to software, hardware and the North American market, while bringing him into the health and fitness space he would continue to focus on.

3. The US Health Market Showed That Small Audiences Can Support High Ticket Prices

Patrick found in North American interviews that consumers of health and fitness products were typically 10–20 years older than comparable users in China. Domestic users, including StressWatch users, were concentrated in the 20–35 age range; in the US, actual buyers were more likely to be 40–50, alongside a large population of older users. He attributed the gap to differences in demographics, accumulated wealth and health-spending habits.
Demand in the US was also more deeply segmented. WHOOP targets obsessive athletes, trading a screenless design for a larger battery, more frequent measurements and longer battery life; the hardware costs roughly $200–300, with another $20 or so charged each month. Oura likewise charges a monthly subscription of roughly $6 on top of its hardware.
Hatch’s bedside lamp, which wakes users gradually with natural light, is another hardware-plus-software-subscription example. The host added that when Americans visit a doctor, physicians often ask whether they exercise and how often; patients can show the data directly, and Apple Health data can connect with the Apple Watch. Patrick believes the high cost of healthcare and cultural differences together increase the value of health data.

4. “100 Bad Ideas” Was Both a Creative Principle and a Two-Year Runway

The studio’s name came from a magician’s story: the source first calls him Blake White and later Blake Wood. David Copperfield asked whether he could guarantee bringing one good idea; he said he could not guarantee even one, but could guarantee a constant stream of bad ideas. Patrick took from the story not false modesty but permission to make mistakes.
He and co-founder Alex initially just wrote ideas in a document. The first line on the cover page read: “OK, let’s write down 100 bad ideas first, then we’ll talk.” After seeing it, a friend suggested naming the studio A Hundred Bad Ideas.
The pair accepted the worst-case scenario of having no income for two years. Patrick had no mortgage, loans, marriage or children, and was willing to move back home to drive burn rate as low as possible; he described having two years to experiment as “a luxury for young people today.”
The time-boxing mindset was inspired by creator Shiufen, who once gave himself six months, listed roughly 20 topics, and planned to return to work if nothing came of them. The team built an AI Bible website and wrote down ideas for an Apple Watch sun-exposure app and a diving tool; the diving project even prompted Patrick to travel to Mexico to study, but ultimately stopped when they could not obtain a diving API that required strict credentials.

5. StressWatch Nailed the Intersection of Wearables, HRV and East Asian Stress

During the pandemic, demand for blood-oxygen monitoring helped drive adoption of Apple Watch and other wearables. Patrick believed that as the installed base grew, a market would emerge for vertical software built around sensor data.
WHOOP and Oura had already established HRV, or heart-rate variability, as an important indicator of physical recovery and mental stress, but Chinese users had little awareness of the concept at the time. Patrick’s insight was that “stress” was also a distinctly East Asian issue: a concept validated overseas was meeting a blank space in East Asia.
Apple’s ecosystem had only a handful of stress apps, while Garmin, COROS and other devices mostly said, “Your stress is 85 today, that’s it.” The team’s third opportunity was to use design to make this male-coded, number-heavy experience more emotional and female-oriented.

6. Emotional Design Turned Stress Data into a State Users Could Relate To

The most prominent element on StressWatch’s home screen is not a number but an animated “stress bubble.” When HRV is high and resting heart rate is low, a green character lounges on a pool float in sunglasses; when stress is high, a red character hides under a tree as wind and rain rage around it.
The environment cycles from clear skies to cloudy weather, light rain and thunderstorms, while the Apple Watch version uses animated expressions to show the user’s current state. Patrick emphasized: “We chose not to say, your stress is 95 or 85,” because an isolated number means little to users.
This framing lets users feel “it understands me” before asking them to understand the metric. The team later added multiple watch faces, combining stress feedback with the Apple Watch’s existing demand for decoration and increasing the product’s everyday visibility.

7. The Team Did Not Force the Product into AI to Chase the Hype

When ChatGPT first took off, the team listed a long slate of AI ideas, but Patrick saw strong Matthew effects in foundation models: unless they could raise the most money and hire the best people, competing with OpenAI in models was not a fit.
If AI becomes a utility like water or electricity, the real asset may be user data—but startups precisely lack user data during cold start. That makes a supposed data moat difficult to establish in the early stage.
The third path was to return to familiar domains and find where AI might fit, but once a real need is found, “whether it’s AI or not doesn’t really matter.” StressWatch ultimately came from the intersection of health, design and Apple Watch, not from trying to satisfy a predetermined AI label.

8. PMF First Appeared as Users Talking on Their Own, Not as a Sudden Revenue Spike

During beta, the team dropped StressWatch into user groups without deliberately operating them. A week or two later, users were still discussing the product every day on their own, matching one of Patrick’s preferred gut checks: “If nobody is talking about it, it’s really dead.”
Early creators were also willing to promote it for free or exchange exposure through redemption codes because they “had never seen a product like it” and respected the level of finish. For a cash-constrained team, creators voluntarily lowering their rates is itself a signal of demand intensity.
More rational signals include growth, conversion, acquisition difficulty and retention. Patrick’s rough benchmarks were around 30%–40% retention on day 1 and 10%–15% on day 7; the numbers are not a universal answer, but they test whether excitement has turned into sustained use.

9. The Biggest Detour Was Treating a Product with PMF as a Lifestyle Business

The team initially believed only Apple Watch users could use StressWatch, so the ceiling could not be very high; 100K–200K downloads would already be a good result. At launch, daily revenue was sometimes only in the tens of dollars, reinforcing the “small utility” thesis.
Apart from fixing bugs and running a few Xiaohongshu and QQ collaborations each week, the team did not commit all its engineering, design and operating resources to StressWatch. It continued exploring diving and other projects, as well as outside work opportunities. Patrick admits that this risk diversification left roughly six months unused.
During that period, some products may have built similar apps outright or added the same stress functionality to existing products, and their growth even surpassed StressWatch’s. Competition ultimately did not become especially intense, with luck playing a role; if he could do it again, Patrick would “not let those six months go to waste.”
His postmortem has 2 parts: concentrate all design and engineering resources on deepening the product, and raise money for rapid growth. The funding need not come from VC, since a utility- and cash-flow-oriented product may not fit venture capital’s return profile; Friends, Family and Fools, a publisher or the team’s own money could be better matches.

10. Small Teams Should Target Needs Big Companies Ignore but Users Cannot Live Without

Patrick uses The Innovator’s Dilemma to explain that the best starting point for a startup is often a countercyclical market too small for a large company to care about. It may not support a big company’s growth targets, but it can be extremely important to a small group of people—“no matter how small the PMF looks.”
In a needs matrix, the 80% strongest needs of 80% of the population—communications, ride-hailing and food delivery—are already covered by large companies. The 20% weaker needs of 80% of the population are better suited as extensions of major platforms; shared bikes, for example, became a Meituan ancillary business that could use the core business’s traffic to lower acquisition costs.
The best fit for a small team is the 80% strongest needs of 20% of the population. WHOOP’s focus on obsessive athletes is a classic example; another is food-photo recognition built only for diabetes patients. Food content may be irrelevant to most people but directly affects health management for those patients.
The 20% weaker needs of 20% of the population historically could not cover a company’s costs, but AI’s boost to programming efficiency may make them economical to serve. Patrick’s priority remains unchanged: entrepreneurs should first find strong needs among a minority, then see whether they generalize outward.

11. “MPF” Means Confirming the Market Before Deciding What Product to Build

Patrick believes the PMF formulation can mislead people into thinking there is a product first, followed by a search for a market willing to buy in. Small teams can reverse the sequence and look for “MPF”: first identify a proven but underserved need, then design the product.
Copying a concept from one country to another is the most direct MPF path. StressWatch was more a case of concept transfer: HRV had already gained recognition in the US, while East Asia remained a blank space, and stronger design lifted an existing roughly 60-point experience to 80–90 points.
For inspiration from rankings, Patrick looks at No. 200–500 in a subcategory rather than the top of the chart. That range means there is “some user demand without a particularly large player”; if people still use a product despite a terrible experience, it may reveal a genuine supply gap.
Being 5–10 points better than a mature competitor is usually not enough, because switching costs exceed the incremental value and existing users instinctively defend their incumbent tool. The host joked about Notion: Patrick finds it complicated, perhaps simply because he is not yet a user; Patrick conceded that another possibility is that his own documentation needs are not especially strong.

12. An Existing User Base Makes It Easier to Derive a Second Product Than an Abstract Technology Trend

Patrick remains wary of the technology-first path: starting from a major AI trend and narrowing down to a specific product often means users appear only at the end. The reverse path is to understand a real group of users first, then find adjacent needs around their characteristics, making cold start cheaper and more tangible.
StressWatch had already amassed a large Apple Watch user base, so when the team built PeakWatch, it could send a survey to screen for fitness enthusiasts and pull them directly into testing rather than finding beta users from scratch.
He cites similar advice from Small Is Beautiful: enter a neighborhood, club or other concrete community, observe what people need, sell that community something it genuinely needs, and expand gradually instead of assuming a grand market from the outset.

13. WWDC Offers a Temporary Platform Tailwind, Not Just a Feature List

After Apple expanded APIs related to mood logging, developers could write data into HealthKit more easily. A mood-tracking app might previously have required its own server; now data could be stored directly in the user’s HealthKit, potentially without a server at all.
StressWatch added mood logging around this change, aligning with Apple’s priorities at the time and earning inclusion in a mood-tracking feature and global promotion. Patrick repeatedly watches WWDC to identify what Apple wants its ecosystem to build next.
Apple Watch’s sun-exposure feature also generated the team’s idea for a sun-exposure product; earlier, the opening of desktop widgets had spawned a wave of theme and customization companies. For iOS developers, each year’s platform changes can reallocate organic traffic.

14. The Growth Window Closes Just Like the Product Window

After StressWatch began growing domestically, the team quickly considered overseas expansion, but lacked the experience and investment to act. It was not until more than a year after launch, around June last year, that it materially increased spending across multiple countries.
In many markets there had initially been no comparable products, allowing the team to secure a position quickly at very low cost. By the time of recording, acquisition costs in some markets had risen 2–3x. Patrick’s conclusion: “Every product has its window,” and growth opportunities do not wait forever.
Large markets matter, but Japan, South Korea and several European countries also have strong purchasing power. A small team can still benefit from first securing a smaller market.
The team’s process has now changed: validate PMF, fix serious bugs, then immediately launch the relevant overseas push. Patrick treats growth as equally important as product, because after going from zero to one, moving from one to 100 requires “100x, 1,000x” the resources to build out.

15. Utility Apps Offer Measurable Payback, but Marginal Acquisition Gets More Expensive

When overseas spending was increased last June, Patrick saw the ad account spend heavily for the first time: “The money spent in a few days might be what I had earned over the previous few months.” Because it was his own money, he slept badly for nearly a month and checked conversion first thing every morning.
The team brought in friends with user-acquisition experience, kept producing creative, and only increased the budget after confirming that the campaign was not losing money and that funds spent would come back after at least two months. Without outside financing, every hiring and acquisition decision had to clear an ROI test.
These apps lack the network effects of internet platforms, so CPI can rise as scale increases; paid acquisition also attracts copycat competition, pushing prices higher for everyone. Strong top-line revenue does not mean margins will naturally improve like those of a platform business.
The host suggested that a large base of long-term subscribers might eventually let the company reduce acquisition and enter a harvest phase. Patrick’s reservation was that users will always cancel. The true ceiling arrives when new users can no longer cover churn, at which point the company can only maintain a smaller budget and a stable user base.

16. The Team Is Deepening 3 Products Instead of Opening Endless New Bets

The team currently operates StressWatch, PeakWatch and a product called “练就,” with nearly 10 full- and part-time employees divided mainly into 2 small teams focused on StressWatch and PeakWatch.
Patrick believes that once a health utility reaches a certain stage, it is counterproductive to keep stuffing in miscellaneous features. Overseas users in particular are accustomed to products that do one thing well. Only after a product matures does the team have capacity to invest in the next R&D cycle.
The team remains open to joint development, product partnerships and acquisitions of small products or teams, but Patrick does not currently want the company acquired. He repeatedly emphasizes that “people matter most”: fully owning StressWatch already takes all his capacity, so each new product needs a new leader.

17. App Factories Have Moved from Organic Distribution to Global Paid Acquisition

The host noted that “factory” evokes assembly lines, knockoffs and emotionless copying. Patrick acknowledged that the market has its share of low-quality operators, but also teams that make every product with care. He prefers to describe his current setup as a “small workshop,” with the goal of gradually building a factory.
The first generation of App factories benefited from the demand explosion after the App Store launched, growing through polished apps, platform promotion and organic traffic. Early utilities such as countdown apps belonged to this era. As supply increased, organic distribution gradually diluted.
The second generation was built on global ad platforms such as Facebook and TikTok. These platforms aggregate users worldwide and let developers reach people with stronger willingness to pay and monetization potential at relatively low acquisition costs. Overseas companies such as Ricky used this infrastructure to access global markets.
The host asked why the first generation did not naturally evolve into the second. Patrick’s answer was simple: “Growth capability is pretty hard to develop.” Experience with product, animation and platform recommendations does not automatically translate into creative production, campaign optimization or cash-flow management.

18. Mature App Factories Do Not All Use the Same Playbook

Patrick admires Beijing-based Underjoy, whose core product is essentially a to-do list but not for conventional productivity users. It targets homemakers, self-improvement and daily routines, packaging cute design, habit formation and prebuilt lists such as “5 habits that will change your morning.”
Me+’s edge lies not only in positioning but also in onboarding and paid conversion: “After you complete the onboarding, you feel, no, I have to pay for this.” France’s Lumi stands out for the quality of its website and product design; Ukrainian company Genius’s BetterMe ranks highly in health and fitness categories across multiple countries.
French hyper-casual game company Wudu splits innovation across 2 organizations: a prototype team experiments with gameplay while the growth team runs initial ad tests. Once the data is strong, the project is immediately handed to the business team to complete development, paywalls and monetization. Patrick sees the setup as “very much like a large incubator.”
Other factories win through portfolio breadth and conversion. AIBY is strong at commercialization; Chinese team ABISHKKING repeatedly repackages the same core product around abs, chest, men, women, 7-day and 30-day goals, at one point accounting for half of the global health and fitness download top 10 for a year during the pandemic.

19. AI’s Real Multiplier Is the Speed at Which Organizations Validate Bad Ideas

Looking back on the past two years, Patrick first says, “I was extremely lucky.” Many projects from the same period have stopped, while StressWatch continued growing and enabled a nearly 10-person team to break even; competitive pressure exists but has not yet become existential.
The bigger personal shift was from product designer to full business owner: asking the questions, hiring, launching projects, choosing products, allocating marketing budgets and facing “how much do we spend and how much do we earn each month.” Commercial outcomes make value judgments “brutal and direct.”
He sees at least several opportunity sets today. Global infrastructure lets developers in Vietnam, Turkey and elsewhere build global factories with dozens of people and dozens of products; even if a team never visits a particular African market, it can still reach users there through that infrastructure.
AI further lowers the cost of multilingual work, user communication, development and experimentation. Patrick brings the conclusion back to the studio’s name: most ideas will not work, but an organization that can experiment faster and more cheaply has a higher probability of finding real PMF—so “100 bad ideas” can become “1,000, 1M bad ideas.”

98. A Paper-by-Paper Guide to Classic Work on Robot Foundation Models and VLA—“Humans Are the Most Intelligent VLA”Vol.165 Macro Talk 81 | The Tariff Shock, and Some Changes in China’s Consumption and Capital Markets (Recorded Apr. 3)