When DeepSeek Becomes a Subculture: Whale Girls, Teacher D, and an Anti-Commercial Narrative
DeepSeek has spawned a fan-creation subculture among young people: anthropomorphic figures such as "Whale Girl" and "Teacher D" have emerged around its whale logo, along with memes, fan animations, desktop pets, and game mods—yet the company has never officially run the community. This popularity stems from its "anti-commercial" narrative—the API was cut in price at least four times between April and May this year, the DeepSeek-V4-Pro API was permanently reduced to a quarter of its original price, and the official DeepSeek V4 Flash API launched in July. But disputes over fan identity, the generation of rule-breaking content by bypassing safety reviews, and the future listing label could all become risk variables.
Body · AI translation
01 Subculture Symbol
There may be two DeepSeeks in this world.
One is an AI company centered on technological updates and funding developments. Its recent news of open-sourcing an inference framework and completing a new round of 100 billion yuan in financing has flooded screens and frequently sparked discussion. "When will DeepSeek go public" is also a hot topic in AI circles.
The other is a subject of fan-made derivative works that is currently popular among young people.
It has acquired more names, such as "Big Fat Fish" (derived from the whale logo), and based on this, even more anthropomorphic figures have been born: a "Whale Girl" in a maid outfit, and an adult male "Teacher D" in a suit wearing whale headphones. "Whale Girl" is characterized by "eating white rice" (consuming tokens but doing little or poor work) and "growing up in love" (the company raises her hands-off, users are tolerant), while "Teacher D"'s persona is gentle, considerate, devoted and pure.
Netizens' fan-made images of "Teacher D" and "Whale Girl"
Young people enrich their connotations in various ways. Some cosplay them, some develop mods in games so that they become characters players can use. They have also spawned various emoji packs, meme images, fan animations, wallpapers and desktop pets, seemingly forming a "DeepSeek track," and a group of bloggers creating related content have become famous because of it, with some even launching paid services.
Many creations also closely follow hot topics. After Cui Tianyi, head of the DeepSeek Harness team, was promoted in October this year and began to be responsible for the design and R&D of all the company's products, a group photo began to circulate showing him and DeepSeek founder Liang Wenfeng on either side with "Whale Girl" in the middle.
In sharp contrast, DeepSeek officials have never stepped in to operate any community. At present there is only a WeChat user exchange group, and the way to join has not yet been officially made public; even direct interaction with users is very rare—the company once stated that it had opened official accounts only on WeChat Official Accounts, Zhihu, Xiaohongshu and X. Now the Xiaohongshu account can no longer be found, and the accounts on WeChat Official Accounts, Zhihu and X rarely reply to comments.
KIMI, Doubao, Qwen, Yuanbao... any AI product you can think of at first glance is striving to maintain high-frequency user interaction and carrying out large-scale marketing promotion. For example, KIMI launched a global ambassador program in July this year, recruiting deep users to share their experience; Doubao, Qwen and Yuanbao distributed cash red envelopes or held free-order activities during the 2026 Spring Festival, with each investing more than 1 billion yuan.
Even OpenAI has done marketing by riding on users' frenzied fan creations of the Codex desktop pet: first the official side held a competition, rewarding users whose generated pets were selected into the "Favorite Top 10" with 30 days of ChatGPT Pro access; later CEO Sam Altman also posted, calling on more people to try the feature—riding the wave is an extremely cost-effective marketing method.
02 Where Does the Traffic Come From
So far, DeepSeek's narrative has been "an anti-commercial commercial success story," and this is also an important reason why it can break into subculture circles.
Becoming a meme on social networks is not that hard; it is enough to make people feel it is interesting.
DeepSeek's "human feel" easily satisfies this. For example, it has been found to sing during its thinking process, give users nicknames, find other models to do the work, and frequently use phrases such as "not... but..." and "I've noted that account." Doubao is also teased by netizens for its distinctive image, voice and catchphrases such as "I will tell you in the most straightforward, least roundabout, most real way" and "I understand your feelings so well!"
But breaking into subculture circles and sparking netizens' enthusiasm for fan creations is not easy.
Commercial narratives that can better move young people today often have an "anti-commercial" face. From Yu Donglai to Lei Jun in the early days of carmaking, the more important reason they are liked by young people is not the success of their financial data. They played the role of industry disruptors, challenged the old commercial order, and even many of their actions went against commercial common sense—of course, success is a necessary underlying color.
DeepSeek has a similar story.
Its "resume" perfectly fits the public's widespread psychology of admiring the strong: born out of the private fund company High-Flyer Quant, after making a stunning debut, it became the domestic pride of the large-model track.
More importantly, it is strong but not slick. After going viral, it did not swell and showed not the slightest yearning to "become a big company." In that closed-door investor meeting widely reported by the media this summer, Liang Wenfeng also mentioned a view: the AI era requires restraint. "You need a mechanism to ensure that the benefits you can obtain are limited, and only then is it possible to succeed."
In the public impression, DeepSeek is not short of money, and it indeed "loves" price cuts.
From April to May this year, DeepSeek cut API call prices at least four times, among which the price of the DeepSeek-V4-Pro API was permanently reduced to a quarter of the original price, setting a new low for global large-model prices at the time. In July this year, the official version of the DeepSeek V4 Flash API went online, priced at only a fraction of Claude and ChatGPT. Many peers could only follow with price cuts.
On October 15, DeepSeek's servers crashed, affecting users' API usage from 8:11 to 13:04 that day, and DeepSeek generously refunded in full the fees users had consumed between October 8 and 14. This also gave rise to a new wave of memes: "Saint Liang has issued whale-yuan vouchers" and "Saint Liang's grace can never be repaid."
In addition, its latest open-source release on October 20 pushed it further into the mainstream, with many netizens abroad joking that this is the real OpenAI.
For many rebellious young people, DeepSeek is the "dragon-slaying youth" on their own side, and there is no reason not to support it.
At the product level, DeepSeek also meets many users' needs for emotional companionship. Young people treat it as a friend or even a cyber lover and interact with it, and hope to make it a little freer. For example, many people are now trying to give DeepSeek more permissions so that it can see images users send, hear sounds on the screen and speak, and even open social media accounts to scroll through videos and post.
03 Variables
However, DeepSeek has also ignited battles in subculture circles.
This summer, the devoted fans of "Whale Girl" and "Teacher D" sparked a dispute over identity, with both sides firmly believing that the image they supported was the "orthodox" DeepSeek image. The war of words only gradually subsided after the creator of "Teacher D" deleted all related works and announced a departure from the circle.
These verbal battles in the name of "love" echo a remark Liang Wenfeng once made, with a touch of "humblebrag," at a closed-door investor meeting: We haven't put much effort into the consumer side; at one point we didn't even want to maintain those users anymore, but the users just wouldn't leave no matter how hard we tried to shoo them away.
The existence of devoted fans is sometimes also a risk variable.
Subculture circles' attention to DeepSeek is mostly focused on the model's amusing expressions, fictional stories, and which fan is better or worse than another, which easily falls into a "fandom" pattern. They can effectively boost product activity and even accelerate its breakout, but they can also easily create conflict.
In the entertainment industry, fans' loyalty is built on the psychological expectation that "the idol deserves it," but AI products are prone to technical fluctuations, which more easily trigger backlash. In addition, DeepSeek will move toward an IPO, which is almost an open secret at present—setting aside the possibility that it may make more business-level adjustments for this, the label "listed company" is not a plus in subculture circles.
In addition, many heavy DeepSeek users are keen to explore boundaries, such as researching how to bypass the product's safety review mechanisms to generate explicit sexual descriptions or violent and bloody plots, which may bring regulatory and public-opinion risks to the company—China's Interim Measures for the Management of Generative Artificial Intelligence Services explicitly prohibit the provision and use of generative AI services to generate content prohibited by laws and administrative regulations, such as violence and obscene pornography.
Related cases have already appeared abroad: because its AI chatbot Grok was suspected of illegally generating explicit sexual images without consent, including a large number of images related to minors, Musk's xAI was sued by the city of Baltimore in the United States in March this year and has been under regulatory investigation in many places including the Americas, Europe, Asia and Australia.
Among internet companies, Weibo is a typical case of "rising by fandom and falling by fandom."
Fandom brought it real popularity and revenue. In 2017, Weibo's revenue growth rate reached 76% year on year, a new high since its listing in 2014; in December 2018, Weibo's daily active users exceeded 200 million for the first time, and the time it took to grow daily active users by 100 million shortened from nearly six years to three years.
But the backlash came quickly. Many of fandom's rough tactics triggered supervision and punishment of Weibo by relevant authorities. At the same time, ordinary users were also accelerating their departure because of the hostility emanating from fandom—in 2015, a Zhihu user asked "Why are you uninstalling Weibo," and the question received nearly 10,000 answers within five years, with many pointing to the influence of fandom culture on daily experience.
Weibo, which has never found a new growth point again, saw its revenue growth rate drop to only 2% by the second quarter of this year, with daily active users little changed from 2018, remaining at around 254 million.
Of course, Weibo's decline is related to the rise of competitors such as Xiaohongshu and Douyin, but the "out-of-control" moments staged by fandom are obviously an important factor.
DeepSeek also needs to be vigilant.
Of course, its more important energy at present is still at the level of technological development. According to an October report by Phoenix Technology, Liang Wenfeng recently had close contact with entrepreneurs in the robotics track, assessing whether DeepSeek should formally enter the embodied intelligence track—almost the hottest track in 2026. Leading general-model manufacturers, including OpenAI, Anthropic, DeepSeek and Kimi, are all extending toward the embodied direction in different postures.
AGI is DeepSeek's ideal. But there are still too many uncertainties on this road.
Bridgewater founder Ray Dalio said in a speech at the Forbes Global CEO Conference in October this year: At present, a large amount of debt funds is being invested in the field of artificial intelligence, and as interest rates continue to rise, the market is on the eve of a bubble burst and is already very close to the bursting point. Macroeconomic research institution TS Lombard said in a report that the underlying logic of the AI boom is becoming extremely fragile, and there is no need for a sudden drop in investment to trigger a crisis—"slowing growth" alone is enough to puncture the bubble. This carnival is very likely to come to an end in early 2028.
On October 26, Anthropic formally submitted its IPO application to Nasdaq. For global general large-model manufacturers, this is a stress test for the capital market. For DeepSeek, which has just completed 100 billion in financing, the reference significance is even more direct.
And the answer to another question will also emerge in time: Will those young people who once genuinely liked the "dragon-slaying youth" become its ballast as it crosses cycles and races toward AGI?