Tianya is officially back today_the Chinese web it hit has been rewritten by AI, but there is one thing it may have bet on

At 0:00 on June 1st.

Tianya Community officially returned after stopping serving for three years.

The domain name has been changed. Changed from the original tianya.cn to tianya.net (the announcement stated that "due to various reasons, the tianya.cn domain name is temporarily unavailable." Hainan International Arbitration Court completed the data migration arbitration award, which can be regarded as legally saving the data).

The first step of coming back to this day is to browse some of the best posts. Interactive functions (posts, replies) are said to be gradually restored within June.

In the public report, reporters from many media went for a trial interview this morning, but they didn't open it for the first hour or two. The experience itself is quite Tianya-come on, but don't try too hard.

Many people will write about this as a nostalgic event. It has been gone for three years, and my old friends have come back. Ghost blows out the lantern, Ming Dynasty events, lotus lotus bunge nonsense, Sister Furong, Kaidi, International Observation, and a series of names have been dropped.

I don't really want to write this way.

Because on the day when Tianya returned, the Chinese web it faced had been completely rewritten by AI. The Internet it knew was long gone.

But the core reason why I want to talk in detail about this matter is that the timing of Tian Ya's return accidentally hit a severely underestimated reverse demand. If it can grasp this matter, it may find a second spring in its AI era. If you can't catch it, there is a high probability of death again.

!

Let's make it clear first, how did Tianya get to where he is today?

Tianya was founded in 1999. The founder, Xing Ming, was born in Wenchang, Hainan Province. He was born in Cangzhou, Hebei Province. He graduated from the Chinese Department of Sun Yat-sen University in 1991. This company started as a stock forum for Xing Ming to communicate when he was personally investing in stocks, but later gradually became one of the largest comprehensive communities on the Chinese Internet.

Its peak data-more than 130 million registered users, more than 250 million monthly users, and once landed on the New Third Board, with a maximum valuation of 1 billion yuan.

In the China Internet content ecosystem of that era, most of the native original content came from this kind of PC forum. Novels such as "Ghost Blows Out the Lantern" and "Those Things in the Ming Dynasty", which were later named "the first works of China's online literature," were all serialized one by one in Tianya. Sister Furong, the first generation of Internet celebrity, came out of Tianya Huo. Lotus has told countless stories that were later adapted into films and television. International observation was the birthplace of the first wave of civil geopolitical discussions in China, and the Economic Forum was the first place where ordinary people could see the content of "How do retail investors and tycoons gamble?"

If you open any public account, any popular Douyin video, or any hot topic on Weibo today, and look for three layers at the source, there is a high probability that it will be a certain Tianya God post from 2003 to 2012.

After this era passed, Tianya failed to catch up with the mobile Internet. Weibo comes out, WeChat comes out, and Short Video come out. Wave after wave of migration, and users stream from PC forums to mobile apps. Tianya failed to keep up, and traffic continued to shrink. Coupled with founder's personal equity and debt problems, company-level litigation, bank loan cuts, and employee departures, multiple difficulties have come together.

On April 1, 2023, Tianya suspended access due to telecom IDC arrears. On May 27, Tianya Community publicly admitted that it had difficulty in liquidity, and that the accumulated arrears in telecom IDC fees in recent years had reached an uncoordinated state.

After the shutdown occurred, founder Xing Ming successively appeared enforcement information on his side-restriction orders on consumption, dishonest executors, final cases, and equity freezing. In October 2024, Xing Ming was restored to execution of 20.66 million yuan.

Song Zheng, one of the co-founders and former executive editor-in-chief, held a "Seven Days and Seven Nights, Restart Tianya" charity sale in the live broadcast room from May 28 to June 3, 2023. After being broadcast live for 7 days and 7 nights, less than 150,000 yuan was finally raised.

This figure is an indescribable comparison for a company that "once valued at 1 billion and registered users of 130 million."

is back this time, how did it come back?

In February this year, the Tianyake public account issued a "Create Together with You, Restart Tianya-Announcement on New Tianya Recruiting Creation Members", recruiting 9999 "Creation Members" and selling for 1999 yuan each. Calculated, the total crowdfunding amount is 19.988 million yuan.

This crowdfunding opened up the situation. In the same period, the "New Horizon Joint Working Group" was established to implement special funds, professional teams, and various preparations needed for data migration, platform optimization, and access restoration.

Regarding data, the Hainan International Arbitration Court issued a legally binding data migration arbitration award to rescue Tianya Historical Data from China Telecom.

Regarding the domain name, due to "various reasons", tianya.cn cannot be used for the time being, so tianya.net. The operating entities are Chengdu Tianyake Network Technology Co., Ltd. and Tianya Haodongshi (Hainan) E-Commerce Co., Ltd., as the "data entrusted operators." The ownership of the original brand and historical data still belongs to the original Tianya Community Company.

Another detail was also added-each Creation member issued a "Tianya Reboot Digital Badge" and forged it on the domestic compliant ant chain. The total number was strictly 9999, with voting rights for community governance and offline cultural tourism consumption rights. This operation has packaged crowdfunding into the dual identity of "digital collection plus founder shareholder."

This is the most poignant and cult "self-help" battle for the Chinese Internet in the past three years. Founder Xing Ming limited consumption and a background of dishonesty. Co-founder Song Zheng raised 150,000 yuan live broadcast in 7 days. Later, each of 9999 old users paid 1999 yuan, which added up to gather the necessary capital.

This process is not glamorous at all.

But it really came back.

But what is most worth discussing about coming back is not nostalgia

To be honest, those Tianya stories from 1999 to 2012 have been written by countless public accounts. Memories kill are easy to write. But with the memories written and the article finished in 5 minutes, nothing can help you judge what the company will become in the next 12 months.

What I want to do is another thing--

See clearly what kind of Chinese web Tianya is facing on the day it returns. Then judge whether there is a place in the old forum form in this new web.

Let's first put out the current status of the Chinese web on June 1, 2026.

You open Baidu, the top half of the page gives you AI answers directly, and the blue link is pushed to the bottom half of the page. Open Quark and default to AI dialogue. Open WeChat and search, and find that they are all public account articles and Mini programs. Most of the articles themselves are written by AI. Turning on Douyin, there are a large number of AIGC videos in the recommendation stream, and almost half of the IP creators use AI to write scripts. Open Weibo, and there are entries related to AI models on hot searches every day. Opening Zhihu, the answer is increasingly like a Wikipedia-style standardized "AI-optimized version". Open the little red book,"AI retouching" and "AI copywriting" have become basic skills.

Overseas, on May 19, Google I/O announced the biggest change in 25 years. The intelligent search box was redone, and the monthly activity of AI Mode exceeded 1 billion. Within a week, DuckDuckGo app installs rose by 18% to 30%(month-on-week), and a group of users are voting with their feet to protest "AI takes over search."

AI has eaten up most of the Chinese web's three links: content production, content distribution, and content consumption.

This is the world Tian Ya will face when he comes back.

It looks hopeless. A PC forum born in the late 1990s is back in an era when AI has been written all over the web-who will go?

But the opposite is true.

!

Why did I say Tianya ran into a reverse scarcity demand

Three pieces of evidence.

** First item, real data from Reddit **

Reddit is currently the largest "real-life discussion" platform overseas. Not only did it not die in the AI era, it became one of the biggest winners.

On February 5, 2026, Reddit released its 2025 Q4 financial report. Look at the specific data--

Search WAU will increase from 60 million in Q4 in 2024 to 80 million in Q4 in 2025, an increase of 30%.

Weekly active users of the "Reddit Answers" section (an AI-driven Q & A product) increased from 1 million in the first quarter of 2025 to 15 million in the fourth quarter, 15 times. In the middle is the quarterly acceleration-Q2 6 million, Q3 7 million, and Q4 15 million.

More critical is its data authorization business-selling real-life discussions accumulated on Reddit for years to AI companies as training data. This item is included in the "Other Revenue" category and will be approximately US$140 million for the full year of 2025 (approximately 33 to 36 million per quarter). Among them, AI data licenses signed with Google and OpenAI account for about 10% of Reddit's total revenue according to industry estimates, with an annualized annual revenue of about US$130 million.

Reddit's total revenue for the full year of 2025 was US$2.2 billion, a year-on-year increase of 69%. Q4's single-quarter revenue was US$726 million, a year-on-year increase of 70%. The daily activity was 121.4 million, a year-on-year increase of 19%.

Reddit CEO Steve Huffman clearly repositioned Reddit at the 2025 Q4 earnings call-from a "social platform" to a "place where people seek answers." He said,"I think there is still a lot of potential for development on this road. We will use Reddit Answers to handle more and more user queries."

This is a company that has been rising all the way with the "real person discussion database" and the dividends of the AI era. What does Reddit's chances of winning come from? From "In the AI era, users are beginning to need to go to places where real people gather to find answers."

** Second, model collapse consensus in academic circles **

In July 2024, Nature published a paper by Ilia Shumailov et al.(titled "AI models collapse when trained on recursively generated data"), which gave a formal definition of "model collapse"-using AI-generated data to train the next generation of models, performance crashes will occur several generations later. The "tail information" of the model distribution is first lost, and then the overall convergence to a point estimate with extremely small variance. Finally, the model cannot be restored to the state of "training based on human data."

Many follow-up papers will be published one after another in 2025. In September, there was an arxiv preprint (No. 2509.16499) titled "A Closer Look at Model Collapse: From a Generalization-to-Memorialization Perspective", which further proved from the perspective of "generalization to memory" that large-scale use of diffusion models leads to the proliferation of AI-generated data and the degradation of recursive training performance.

In February, a paper titled "Machine-generated text detection prevents language model collapse" was published, demonstrating that machine-generated content detection and human data protection are the core defense against model collapse.

The Nature article has been cited more than 1000 times in the industry (according to Google Scholar). The word "model collapse" has entered the basic understanding of the AI industry from purely academic circles.

What does it mean? This means that AI companies are beginning to regard "real-life original content" as a strategic asset. Reddit's data empowerment revenue is just the beginning. In the next 12 to 24 months, any platform with "real-life in-depth content" will be pursued in reverse by AI companies as a "data gold mine."

** Third, AI companies are publicly stealing real data **

This is already happening, not a prediction.

Reddit signed a $60 million annual data licensing contract with Google (public in 2024). During the same period, an agreement was also signed with OpenAI that did not disclose the specific amount.

Stack Overflow (Programmer Question and Answer Community) has signed a partnership with Google to allow Gemini users to directly access Stack Overflow data (disclosed in 2024). The specific amount was not disclosed.

The Wikimedia Foundation (Wikipedia's parent company) publicly stated in 2024 that it was negotiating licensing with a number of AI companies-its request was that "our content will be crawled away by you to train your models, and you will have to at least pay the corresponding infrastructure costs."

The New York Times sued OpenAI and Microsoft in December 2023, accusing them of using millions of New York Times articles without authorization to train GPT models. This case has not yet been closed, but it has set a legal precedent for "copyrights of real-life original content demanding compensation from AI companies." Once the verdict is in New Zealand's favor, it can be invoked by all similar platforms.

Similar actions in China have not yet been made public on a large scale. But at the end of May and early June, DeepSeek's 70 billion financing is coming to an end and Anthropic's 965 billion valuation overtakes OpenAI, which has already sent a clear signal to the industry that the growth of big model companies from now on will not depend on "training a bigger model"(That road, Llama 4 has 30 trillion tokens at a time and DeepSeek V4-Pro has 33 trillion tokens, which has pushed the data scale to a new benchmark), but relies on "getting exclusive, high-quality, uncontaminated real-life data."

On this road, whoever has Reddit's "20-year-old database of real people's violence" will have leverage.

What do these three items add up to?

The day when Tianya came back--

On the one hand, Reddit relies on real-life data to increase weekly activity by 30% a year, Other Revenue is US$140 million, and Steve Huffman repositioned as an "answer place."

On the one hand, it has become a consensus in academic circles that AI training must retain real data.

On the one hand, AI companies openly snatch real data, ranging from Reddit's 60-million-year contract to New York's lawsuit against OpenAI.

Put the three together-** Real in-depth content is a scarce product in the AI era, and this scarcity will become more and more serious. **

This is exactly what Tianya used to be best at-slow pace, long posts, multiple opinions, and real people, and a post of thousands of words can be discussed for months.

Old users of Tianya like to complain about "Why is Tianya's reply so slow","Why is Tianya's post so long" and "Why is Tianya quarreling for several years"-this is a shortcoming in the PC era, but it is a scarce resource in the AI era.

This is what I said,"Tianya has hit a reverse scarcity demand." Its old form (PC forums, long posts, slow-paced, in-depth discussions with real people) is a backward product in the mobile era from 2013 to 2023, but is scarce in the 2026 AI era.

But whether you can catch this matter is another matter

Back to reality--

Founder Xing Ming limited consumption and a background of dishonesty. The co-founder raised 150,000 yuan in 7 days and 7 nights, and 9999 Creation members raised 19.98 million yuan. This cost is an order of magnitude worse than a company like Reddit that is already listed on the U.S. stock market, has annual revenue of more than US$2.2 billion, and has a complete product iteration team.

Whether Tianya can seize the window of scarcity of in-depth content for real people in the AI era depends on several things.

** First, can it think clearly about building a "real-person discussion protocol layer" and not copying the PC forum model? *

If we just move the best posts from 1999 to 2023 back to the PC webpage, give old users a nostalgic look at them for a few months, and then copy the PC forum model of that year to continue operating, there is a high probability that they will die again within six months. Because today no one wants to open a browser and go to a PC forum to sit down.

It must reposition itself and change its attitude from "returning to old forums" to a hybrid protocol layer of "in-depth discussion with real people in the AI era plus source verification and training data." Mobile phones first. Community governance tokenization (digital badges have taken the first step). Open data authorization (refer to Reddit model).

** Second, can it get a data authorization contract from the AI company? *

Reddit's US$60 million a year comes from the foundation of Google's data authorization. It has signed agreements with model companies such as Google, Anthropic, and OpenAI. The data API interface is complete and the compliance framework is clear.

Tianya historical data is one of the largest in-depth content assets of the Chinese web. In theory, China's big model companies (Alibaba, Byte, Smart Spectrum, DeepSeek, Tencent, Baidu, Kimi) should all want it.

However, whether it can be signed depends on Tianya's data integrity, data compliance, and legal capabilities. These three things are completely different from "PC forum operation". Whether the current team configuration of the New Horizon Joint Working Group can be handled is a question mark.

** Third, can it attract "New Horizon Generation" users? *

Old users come back nostalgic, which is a bonus of 0 to 3 months. If no new users come in after three months and no new in-depth discussions are produced, the historical data from 1999 to 2023 will be a one-time asset and will not grow on its own.

The key to attracting new users is to provide an experience in the AI era that other platforms cannot provide. Weibo cannot give "depth of long posts", Zhihu cannot give "real people tear", public accounts cannot give "coexistence of multiple views", Douyin cannot give "slow pace", and Xiaohongshu cannot give "in-depth talks on sensitive topics."

If Tianya can establish differentiation in these niche, even if it only catches 1 million core active users, it will be enough to support its position in the AI era. But if we can't catch any of them, then this restart is a doomed nostalgic livestream.

Other "real-life content platforms" in China are also being reassessed

This matter is not just about Tianya's family.

Zhihu has been severely eroded by AI content in the past two years, but its "Real Professional Answer" library is still one of the largest high-quality real person discussion data on the Chinese web.

At Station B, the bullet screen and comment area lengthen the video for in-depth interaction, and the density of real discussions is extremely high. AI companies have great potential as data sources.

Douban, a group ecology plus film reviews and book reviews plus long posts from real people, is the purest record of the "spiritual world of real people" on the Chinese web.

Xiaohongshu, real people planting grass and real life sharing, UGC is highly realistic, and its reverse value has begun to appear in the AI era.

Private communities (WeChat group, public account message area), nail group, flying book group, Discord Chinese version, and various "real people small circles" in the WeChat ecosystem are also quietly emerging.

This wave of return of the world will likely become one of the triggers for the revaluation of the entire "real-life content platform." You may see in the next 6 to 12 months--

Knowing that it has regained the attention of AI companies (data authorization opportunities), the stock price of Station B has been repeated (the value of bullet comments and comments has been recalculated), and Douban may be acquired by an AI company in China (data assets plus user density).

This is a revaluation line that many people ignore in the AI era.

QKPFX7 What does QK mean to ordinary people

After talking so much about the company level, when it comes to you, this matter means three things.

** First thing, if you make content **

Comparing production speed with AI is a dead end-you write an article for 4 hours, AI writes an article for 4 minutes, and you will always lose.

But if what you do is "true position" plus "true personal experience" plus "true discussion participation" plus "true point of view refutation", AI cannot do it in the short term. Even if AI imitates your tone and style, it does not have your "real interests","real pain points","real costs" and "real interpersonal relationships."

This differentiation will be re-priced by the market starting in the second half of 2026. Ordinary "AI-optimized copywriting" will become cheaper and cheaper, and real-life in-depth content will become more and more valuable.

** Second, if you want to find reliable information **

Don't believe all the answers given by AI.

The ceiling of AI answers is the "median answer"-it combines everything on the web and gives you an answer in the form of the greatest common denominator. This answer is sufficient for "factual inquiries"(photosynthesis equations, dates of historical events, code syntax).

But for "judgment + experience + source verification" queries, you must go to a place where real people discuss it. Tianya (if you can get up), Reddit, Hacker News, Zhihu specific big V, Douban specific group, WeChat specific circle, these are things that AI cannot give you.

"Two-Line Information Sources" will become standard for the next 12 months. AI gives one version, the real person community gives another version, and you cross-verify it yourself.

** Third, if you start a business or invest **

"Real people discuss protocol layers" is one of the directions that has been seriously underestimated in the AI era for 12 to 24 months.

Reddit is running overseas. The Chinese web is basically blank. It is not certain whether Tianya can escape, but the company will definitely escape in this direction. There is a high probability that it will be a new company rather than an old platform.

From the perspective of investment, the next few signals worthy of attention--

China's major model companies have begun to publicly announce "data authorization" services (similar to the one Reddit signed with Google and OpenAI). The "Private Domain Discussion Community" tool in the WeChat ecosystem has begun to take financing actions. Chinese long-tail community platforms similar to Discord began to appear. The valuations of old platforms such as Zhihu, Station B, and Douban began to be recalculated.

Last

Tianya officially came back at 0 o'clock today--

On the surface, it looks like a nostalgic story. The old Chinese forum, which was closed for three years, finally came back after the founder's limited consumption, 150,000 live broadcasts for seven days, and 19,999 old users 'crowdfunding.

Digging down a layer is the first time in the AI era that the Chinese Internet ecosystem has shown a signal of "reverse scarcity of the old form." Slow pace, long posts, multiple views, and real people tear these "shortcomings" of the PC era have become hard currency in the AI era.

Digging further down, it is the line of AI training data, data authorization monetization, and scarcity of real-life content that has begun to enter the public eye. Reddit has paved this path overseas-140 million of its $2.2 billion in revenue a year comes from data licensing and other businesses. The Chinese web is basically blank, and a new company will definitely be launched within 12 to 24 months.

I am personally not very optimistic whether Tianya can survive and find a position in this window (the founder's background, team allocation, and capital amount are all too far behind overseas benchmarks).

But the fact that Tianya came back is much more important than Tianya's own destiny--

For the first time, it put the matter of "the reverse scarcity of real-life in-depth content in the AI era" on the table of Chinese Internet discussions.

This will change a lot of things over the next 12 to 24 months. It will change the data procurement strategy of AI companies, change the competitive dimension of content creators, change the valuation model of investors, and change the habits of ordinary users to find information.

As for Tianya itself, its biggest credit may be just to "come back at the right time" and make everyone realize the value of "real-life content" again.

As for whether it can enjoy this wave of dividends itself or pass on the torch to the next younger company, that is another story.

But the torches were already lit.

!