news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

the ai ​​era has arrived, coupled with the economic downturn, more and more independent developers are eager for stories of arr millions of dollars.

and now, this low-probability event is happening.

a few days ago, when i opened x, i was flooded with cal ai messages. founder zach posted that cal ai's revenue exceeded one million us dollars, and the day he achieved this achievement happened to be the first day of his high school. he sighed, "is this fate?"...

looking at the data comprehensively (the total transaction volume included is slightly lower, at us$840,000), cal ai achieved this result in less than 4 months of operation. in addition, it is mainly developed and operated by 2 teenagers and 1 university graduate. one can't help but wonder how the toc ai product, which has been criticized for not making money, can become so successful.

cal ai's turnover changes from the beginning of 2024 to date | image source: diandian data

3 people, $3 million arr

cal ai is simply an app that scans food calories. users take photos of food with their phones, and the app analyzes the protein, carbon water and fat content in the food, ultimately achieving the user's goal of managing weight. as mentioned above, only three people are developing and operating this product. zach is the ceo, henry is the cto, both are teenagers, and there is another person in charge of growth.

cal ai’s onboarding process for new users, which explains how to use the product, the value it provides, and how to achieve the goal | image source: cal ai

after entering the app and completing the new user guidance process, the app will ask a lot of questions (including age, gender, weight, height, etc.) to create an atmosphere for you to create a weight loss plan, and then jump to the subscription page. you can't use it without paying. such an app that must be paid to use maintains a dau of 40,000. according to the existing data, the arr of this app is likely to reach 3 million us dollars.

looking at the achievements of cal ai, zach himself summarized two key points:

first, find a real problem to solve.

scan calories - the main player in weight management | data source: diandian data

based on the above table, calorie scanning -> weight management is a long-standing demand, the problem is real enough, and the ceiling is high enough. the highest turnover can reach tens of millions of dollars, which is on par with heavy games.

from the perspective of online time, there are three main batches: 2009, 2014, and this time, from the end of 2023 to the first half of 2024. except for cal ai, the latest batch of three players are not very strong, with downloads and revenues in the single digit (w).

as a latecomer, zach believes that, with the top players already fairly solidified, new marketing methods can be used to distribute new ai products to a group of users that have not yet been covered, which points to zach's second key experience. he believes that on social media, especially tiktok, it is possible to cause viral transmission without relying on big kols.

zach posted that the video they released has received 3 billion views on social media | image source: zach x account

the key third party, continuously creating popular ai apps

when trying to understand why cal ai was able to achieve such good results, we discovered the founder of another product that may be more familiar to chinese overseas practitioners, brake, who is the third person in charge of growth at cal ai.

brake had just graduated from college. because he didn't want to follow the usual path of working to earn money to support his family, he started learning programming from ai a year ago. it was this person who didn't know code before, relying on chatgpt, he made two million-dollar arr apps. the first one was rizzgpt, which was later renamed plugai. the second one is more familiar to everyone. umax ranked 44th in the latest top50 genai app (by mau) list released by a16z in august 2024.

umax, an ai app that helps male users become more attractive, ranks 44th on the latest top 50 genai app list (mau) released by a16z (umax was launched in january 2024, and its mau in august was close to 2 million, a slight drop from the peak of 2.4 million. but as an app that helps men become more attractive and is not so in demand, the decline in revenue is more obvious. monthly revenue dropped from $330,000 in february to $170,000 in august, almost halving. but the historical cumulative revenue is close to $2 million (the self-announced revenue is $3 million+))

in the cal ai project, zach called brake a cofounder. this was the third ai product he participated in, and he mainly provided advice on growth strategies. in a podcast, brake said that he made money from three consecutive ai apps, following the same "template" and "fine-tuning" each case according to the specific situation.

No.1 PlugAI

brake found that his college roommate often asked how to reply to messages on dating apps, so he found two people to work together to create rizzgpt, positioning it as an ai "dating coach". users upload chat screenshots, and then the app will return replies. the cold start relied on two youtube channels of niche creators (rizz content) without showing their faces. each channel cost $50, and it received 200,000 downloads in less than a week. in 4 months, it received 1.4 million downloads. rizzgpt quickly rose to $200,000/month, and was later renamed plugai.

No.2 Umax

brake left rizzgpt and retained his equity. he thought that the previous success could be replicated and began to think about project ideas. he found a large number of male users on reddit and tiktok who were "looking for transformation" (lookmaxing) and found that this was a long-term phenomenon. he launched a lookmaxing product called umax. the core loop is to download the app -> scan your face -> get analysis results and suggestions.

in terms of startup, we chose to cooperate with the corresponding look maxing kol sam zia to promote the app. the online video won hundreds of thousands of views and the turnover in the first month was 100,000 us dollars. brake also joined an accelerator with umax. after a few months, it reached an mrr of about 500,000 us dollars (diandian data shows that the mrr is lower).

No.3 Cal AI

find zach on x and work with them to continue to verify the "success template". cal ai was launched around may, and it also distributes the app through content creators, and then advertises videos with good results. the team basically looks for creators on tiktok who have weight management needs, such as fitness, health, and even beauty bloggers. the official account has reposted nearly 50 creators' videos, with the number of views ranging from thousands to millions.

templates distributed through trend discovery (reddit, tiktok, instagram, google trends) + influencers, brake has completed the launch of 3 products and achieved rapid revenue growth. however, as mentioned above, when facing different products, the ideas of distribution have commonalities and also have fine-tuning. brake and zach summarized some of the noteworthy details.

1. develop a strategy in advance and estimate cpm

“when you can estimate a profitable cpa and the conversion rate that 1,000 views can bring, you can get a cpm target, which can help developers formulate an effective content marketing strategy in advance. when selecting specific creators, you can roughly estimate the average number of views, which will be closer to the median of their accounts.”

2. prefer to choose underrated creators

some creators have millions of visits, but because they do not look professional enough, they cannot get business orders from brands. in the eyes of the cal ai team, this is a gold mine. in order to further observe and select creators, they suggest observing what kind of users the creators' content will convert and what content in the comments can be used as a standard for good interaction rate, so that users will trust the products they recommend.

"in my experience, most content creators have a negative roi, but some can achieve a 10x roi," brake once concluded. "when i was working on umax, i was very immersed in it, thinking about what information i would want to know and what i would want to hear if i wanted to become more attractive. i looked at the creators in this way, and then he emphasized that the cooperating creators should have entrepreneurial spirit, so as to achieve the highest roi, but most creators do not have it."

3. different cooperation ideas for different tracks

one point worth noting is that when working on umax, brake chose partner with kol, while when working on cal ai, they were completely paid influencer marketing. brake believes that this is due to different strategies in different fields. the field of cal ai is actually weight loss, which is a very large field with many leading products. look maxing is another matter, a very small segmented track, which can be driven by a leading kol.

4. in addition to content creation, creators should also do other things well

for example, copying videos with good effects, adding a referral code system to creators to improve the effect, and running advertisements at the same time will result in a higher roi.

cal ai advertising valuation in different channels | image source: guangda university

cal ai’s download curve | image source: diandian data (cal ai’s budget for tiktok may be the smallest, but all three waves of growth were invested in tiktok (orange))

small teams follow up, and “quick applications” become a trend?

with the success of cal ai, umax and other products,discover trends, create single-function products, and then use social media to promote them and complete recyclingthis whole set of operations has been regarded as a "golden rule" for developers to get rich quickly through consumer-grade ai applications. and it has derived an upgraded version, which is to speed up the launch of new products and rapid verification. a "fast application" trend corresponding to fast fashion has begun to emerge in overseas markets.

image source: x

the first application we noticed was rayz, which is an app that guides users to "tan". users only need to upload a selfie, and rayz can formulate a safe and efficient tanning plan based on the user's skin condition and the weather information of the location. similar to the process of umax discovering the demand, there were also a large number of users who shared their tanning experiences and beautiful photos on tiktok. after rayz was built, it also chose tiktok as the first frontier for promotion. x blogger "guillaume" introduced that rayz has accumulated 40 million views, 100,000 monthly downloads and 80,000 us dollars in monthly turnover on tiktok.

rayz app store poster

what is even more amazing is that rayz most likely achieved the above results with zero promotion budget. we were unable to find any records of rayz's advertising through third-party data platforms, and the three tiktok accounts that brought it tens of millions of views were all created by the team themselves. seeing this, you may be impressed by the rayz team's superb material capabilities, but we have to introduce one of their "incredible operations", which is to directly download beautiful girls' tanning videos and add product screen recordings to turn them into their own materials. through such batch operations, rayz successfully ran out of explosive materials. currently, the top two videos with the most views under the #tanning tag are rayz advertisements, with hundreds of thousands of likes.

tiktok account created by rayz team

rayz is not the only work of this team. after launching rayz in july, they used a similar method to promote demure, a female beauty rating tool, in august, and one minute for god, a religious app, in september.the speed has been achieved to produce one model per month.behind each app is an insight into tiktok trends. demure was inspired by the popular video on tiktok about "ugly facial features and overall harmony", while one minute for god is targeted at religious accounts on tiktok. "one minute for god?" is usually used as the opening line in the video. you may wonder if it is necessary to make a product. the answer of the rayz team may be to leave it to the market for testing. they focus on rapid construction and low-cost promotion to gain more trial and error opportunities.

popular videos on tiktok about "ugly facial features but overall harmony"

tiktok's "one minute for god?" popular videos

final thoughts

under the dual influence of the economic downturn and aigc lowering the threshold for entrepreneurship, a large number of independent developers have been seeking to make apps or web products and earn us dollars overseas in the past year. however, from the community i joined, the most difficult thing for chinese developers with the engineer bonus is to find demand and grow. these developers are not short of technology, but they lack a keen sense of market demand and growth experience. i also hope that the experience of cal ai can give you more inspiration.

back to these products, ai wingman, male appearance enhancement, tanning guide, these products that we initially judged as pseudo-demands, seem to have lived longer than expected. although the user and flow data of the product have declined, at the practical level, demand discovery + fast payment verification "defeated" the arrogant assumption. and cal ai, which absorbs user data and is used for long-term weight management, may last longer. rayz and his team showed a side with a high probability of failure when looking for market demand, and using speed and efficiency to approach "right" is the answer they gave.

brake believes that with the advancement of ai, the cost and time of building ai products will become lower and lower. the difficulty lies in how to bring it to the right people in the right way. under this trend, "quick applications" may become popular, but success is still not easy. the high school student founder of cal ai released 5 products before cal ai, but they did not make any splashes. brake slept on other people's sofas before plugai made money.

all i can say is that ai products are already very competitive, but opportunities still exist if certain conditions are met.