What high achievers and programmers care about is often different from the big things actually moving in the market. Starting with ByteDance's playbook, this article explains why interest should not choose the market, where to look for topics, why behavior matters more than narratives, and where the real incremental opportunity lies in the AI era.
I have recently figured out a few things and want to write them down.
First: what high achievers and programmers care about may be completely different from the big things actually moving in the market.
Second: I suddenly understood why ByteDance became so powerful.
Founders are better off not accepting business interviews casually, and not explaining their logic publicly.
The reason is simple: people who follow business, operate companies, or invest may notice a big thing outside their interests only because of one founder interview. Once you explain it, you have handed the map to someone else.
Fewer interviews really can reduce imitation by competitors.
The public image of TikTok and Douyin is mostly consumer resentment: they waste time, the content is low quality, and the algorithm is addictive.
That impression is a kind of camouflage. While everyone is criticizing them, few people seriously study how they make money or grow.
When choosing a topic, pay less attention to what the tech circle discusses and more attention to:
The tech circle discusses what programmers find interesting. These places show where ordinary people are spending time and money.
“Combine your interests and start from what you enjoy and are good at” is a 100% false concept.
What makes money and what is growing are a systems problem. The right order is to learn what makes money first, then move some of your attention there, experience it, and build intuition.
Choosing a market from your interests lets your attention distribution make the business decision for you—and attention distribution is exactly what is least reliable.
People who succeed through their interests certainly exist, but that is survivorship bias: their interests happened to overlap with a large market. It does not prove the method itself works.
After you know what makes money, move your attention there and experience it.
The purpose of experience is not to decide whether something is good. Personal taste has no vote here. Experience is for learning the user’s language and format:
Once you learn these things, the page and product can feel right to the audience.
Interest has not disappeared. Its object has changed.
What genuinely interests me now is not a particular content field, but finding where the money is and then taking it with a system. That interest is sustainable and can be used in any market.
Many people inside ByteDance were probably like this: they did not necessarily enjoy watching Douyin, but they loved the game of growth. Zhang Yiming himself was not necessarily a typical Douyin audience member.
The two ways of finding demand produce completely different results:
Not being the audience is an advantage. You have no strong likes or dislikes, so you will not abandon something because “that is stupid”, nor overvalue it because “that sounds fun”. Looking only at data makes the judgment cooler than the audience’s own.
Zhang Yiming himself was not necessarily a typical Douyin audience member. He came from a programming background. Douyin was not discovered through his personal taste, but through a system that tested its way forward.
He did not need to be a typical audience member, but he did need to understand audience behavior. He recognized the opportunity through technological change, real consumption data, and the company’s capabilities—not just through his own interests.
| Time | Stage | What happened |
|---|---|---|
| Around 2005 | Graduated from Nankai University in software engineering | Programming background |
| 2006–2008 | Kuxun | Technical lead for vertical travel search; learned how information finds people |
| 2008–2009 | Briefly at Microsoft, then joined Fanfou | Encountered social and feed products |
| 2009–2011 | 99fang | Built real-estate information products and several mobile apps; began to see mobile as the major direction |
| March 2012 | Founded ByteDance | Started in a residential apartment in Beijing’s Jinqiu Jiayuan |
| 2012 | Neihan Duanzi, Gaoxiao Jiongtu, and others | Tested the water with low-barrier entertainment apps |
| August 2012 | Toutiao launched | Used recommendations instead of editors to distribute information |
| 2016 | Huoshan Video, Xigua Video, and Douyin (September 2016) | Bet on several short-video products and let them compete internally |
| 2017 | TikTok launched overseas; Musical.ly acquired in November | Bought a validated overseas model |
| 2018 | Douyin took off | Scaled paid acquisition, variety-show sponsorships, and celebrity participation |
He saw the conditions of production and consumption changing together. In a public speech in 2016, he broke the short-video opportunity into three links:
| Link | Change he saw at the time |
|---|---|
| Production | Smartphones and editing tools let individuals make attractive videos |
| Distribution | Content was multiplying, creating a need for recommendation algorithms to match videos with people |
| Consumption and interaction | Phones moved viewing into fragmented moments, making watching, sharing, and interaction possible anywhere |
This judgment did not require him to enjoy singing and dancing. It required him to understand that behavior once difficult to scale had become convenient and cheap.
He had a window that ordinary founders did not: Toutiao’s real data. According to his September 2016 speech, Toutiao began emphasizing video in May 2015. At the time it reached one billion video views a day, more than the combined reading volume of articles and image galleries. He also said demand was growing quickly while supply had not caught up, and even rough UGC could receive high view counts.
That signal mattered: users were proving demand through behavior, while supply still had obvious room to improve. He was not guessing that young people should like short video. He could observe that when given video, users really watched it, and the scale was growing.
He then judged that ByteDance could capture the opportunity. Short-video growth did not mean every company could win. ByteDance already had recommendation technology, content operations, and user-growth experience from Toutiao. The logic was: people were increasingly willing to watch video, individuals were increasingly able to produce it, and ByteDance was good at sending content to interested people. (This is a summary of the strategic logic, not a direct quote.) It explains why the market had value and matched ByteDance’s capabilities.
The specific product came from team exploration, not from one founder designing it by taste. Public reports show that ByteDance repeatedly discussed short video and later launched Huoshan Video and Douyin; Zhang Nan and Ren Lifeng were also important members of Douyin’s founding team. The more accurate view is that Zhang Yiming judged the direction and organized resources, the team explored the experience, and market feedback decided investment.
This avoids hindsight mythology: seeing the short-video opportunity did not mean he had already foreseen live streaming, e-commerce, and the entire scale of today’s business.
The lesson is to ground market judgment in other people’s behavior. You can dislike a type of video and still observe who repeatedly makes it, who shares it, who pays for it, and whether technological progress has made an expensive need available to everyone for the first time.
Big-trend judgment: technological insight. He saw “phones replacing computers, recommendations replacing search and portals”. That is a programmer’s judgment and the part he discussed most publicly. It tells you where to go, but not what content to make.
Testing demand with low-quality apps: data. Neihan Duanzi and Gaoxiao Jiongtu tested what people really watched and shared. He did not have to like jokes; he only had to look at the data. Neihan Duanzi was later shut down by regulators in 2018, indirectly showing its scale at the time.
Douyin was copied: competitive intelligence. Douyin’s prototype is widely considered Musical.ly, a lip-sync music short-video app that first became popular among American teenagers. Demand was validated elsewhere first. ByteDance saw it, built a Chinese version, and later bought the original outright.
Competition between products let data choose the winner. In 2016 three short-video products ran at once: Huoshan for lower-tier markets, Xigua for horizontal mid-length video, and Douyin for young people in first- and second-tier cities. There was no predetermined winner; the products ran in parallel and the one that took off received more investment.
Only after choosing a direction did he force himself to experience it. Reports say he required executives to make Douyin videos and do push-ups if their likes were below the target. This matches the earlier point: learn what makes money first, then move part of your attention to experience it. Experience comes after judgment and is for understanding the user’s language, not deciding whether the market is worth entering.
Finally, money and machines amplified it. Paid acquisition, variety-show sponsorship, celebrity participation, and an advertising sales team turned Douyin from “taking off” into a dominant platform. These were also the parts he discussed least publicly.
Even Douyin’s founder and executives needed a “punishment mechanism” to make themselves use it. That shows it was not naturally inside their attention range.
If Zhang Yiming had continued working at another company and another company had built Douyin, he might not have spent his spare time scrolling it:
The same person sees a completely different world in different positions. What lets you see Douyin is not personality or taste, but where you stand.
Without a competitor-collection system, you are like “Zhang Yiming working at another company”: you may never see this kind of demand.
The push-up rule also shows that a non-audience can learn. “Scroll it when you have time” will not happen; attention naturally returns to comfortable territory.
You can copy the mechanism and set a forced action:
The second does three things at once: it forces you to experience the need as a creator, produces a sample for the page, and becomes an acquisition channel. This is your own “make a Douyin video and do push-ups”.
Personal consumption preference, understanding user demand, and judging business value are three different abilities.
Liking a product does not mean you know why it can make money. You can rarely use a product yourself and still understand through observation, experience, and data why many people need it. A founder’s opportunity set can be much larger than their own lifestyle.
“Would I use it?” is one sample. “Who uses it, why do they return, and how large is it?” is the market question.
That does not make personal experience unimportant. On the contrary, when you are not a typical user, you need to add experience deliberately instead of letting unfamiliarity judge the market for you.
Market intelligence or competitor signal → business hypothesis worth researching → personal experience and user research → opportunity validation → investment.
You did not care about a demand at first, but noticed people buying repeatedly, competitors growing, and users sharing voluntarily. Then you make, watch, and research. Your experience answers specific questions:
This starts differently from “I like something, so I found a startup opportunity”. Interest-driven work experiences first and searches for an opportunity afterward; opportunity-driven work sees the signal first and adds experience afterward. Both can work, but the latter lets you explore markets outside your lifestyle.
Making personal interest the only entry to opportunity discovery restricts entrepreneurial vision to your lifestyle. Your age, job, income, and social circle affect which needs you encounter, but the market is much larger. Separate these abilities:
Interest can be an excellent starting point, but it cannot define the market’s boundary. “I do not like it, I do not use it, and nobody around me uses it” do not disprove a business opportunity; “I really like it” does not prove it either.
Let market signals decide what to research, and let the research question drive personal experience.
You do not have to become a certain kind of consumer before you are allowed to discover that market. You can first see that people pay repeatedly, then learn why they buy, where supply falls short, and where you could enter.
Here is the useful boundary: your time is limited, so you must filter what to research. The filter should be business evidence, not only personal preference.
Your lifestyle can then be a window onto the market instead of a ceiling on your opportunities.
Some programmers are technically strong but work on needs with low business value. People observe this and conclude that these programmers are trapped in a status competition and local optimization; they should leave the tech bubble, have a life, make more friends, and cultivate more interests.
I think that solution is wrong. Finding demand in everyday life is inefficient and random. The real solution is to build a market-intelligence system that continuously monitors information and shows which competitors are making money.
Leaving the programmer bubble is useful, but some versions of the advice fall into another trap. Choosing low-value demand is a market-judgment problem; making more friends and developing hobbies does not directly solve it. Explaining a business-topic problem as a lifestyle problem can send the advice in the wrong direction.
Leaving the programmer bubble matters because you need broader coverage of demand and business information. A richer social life may produce new clues, but clues are random and may not contain payment intent, acquisition cost, or profit. Labels like “status competition” and “local optimization” cannot replace business analysis without a concrete mechanism.
A market-intelligence system is more direct, repeatable, and accumulative. It lets you search actively instead of waiting for everyday life to trigger inspiration by chance.
The system must answer three different questions:
| Layer | Question |
|---|---|
| Discover markets | Which products show signals of growth, payment, and sustained demand? |
| Judge value | Where does revenue come from? How long can demand last? What are cost and competition? |
| Judge your opportunity | Which users can I serve? How will I acquire them? Why will they choose me? |
“Knowing which competitors make money” is an important entry point, not a complete conclusion. High traffic does not equal money, revenue does not equal profit, and a profitable competitor may depend on a brand, channel, or resource you do not have. The system should treat these signals as evidence to validate, not as an automatic reason to launch.
Personal experience and user conversations now have a clear place: the intelligence system first filters markets worth studying, then you experience, interview, and run small paid validations with questions in hand. These activities explain demand and validate the opportunity; they do not have to discover an entire market at random.
Social life and interests can provide supporting clues, but treating them as the main solution is not direct enough.
Discount founder interviews, retrospectives, and “my methodology”. People usually tell the presentable parts, often choosing exactly what programmers like to hear.
Watch behavior, not narratives. Which pages and terms drive another site’s growth, how it prices, where the money comes from, and where it advertises—these behavior data cannot be faked. That is the value of a competitor-collection system: it bypasses the narrative and looks directly at behavior.
The same applies to your own work. The pages that make a site run may be a few specific types, not the “technology” it talks about publicly. Public messaging is one thing; internally, you must know where the money comes from.
From 2005 to 2015, the mature PC internet and early mobile internet, leading internet companies in China, the United States, South Korea, Russia, and Japan all formed around search, social, and e-commerce.
In the mobile era, the pattern became messaging and social, recommended content, instant transactions, and local services, with operating systems and app stores as an additional layer of entry points.
Two capabilities created new forms:
Content platforms do not only absorb entertainment time. They increasingly handle product discovery, experience search, and purchase decisions. The boundaries between content, social, search, and commerce therefore begin to overlap.
The core increment of the AI era is lowering the cost of “understanding, creating, and executing”, so more people can obtain results directly from intent.
New supply includes images, video, music, design, game assets, and software. Needs once abandoned because people could not make them, or because they were too expensive or time-consuming, can now become real consumption.
But more generation does not mean the same increase in business value. The easier content is to produce, the more important selection, taste, credibility, and distribution become.