A 20-year-old founder hires an 18-year-old employee and receives investment from a 19-year-old.

null

Author: LatePost

A 17-year-old AI intern earns 5,500 yuan a day, considered a "mid-level" in 1998

A well-known venture capital firm established over 15 years ago held a dinner, inviting select guests who weren’t in suits but mostly wore black, white, or gray T-shirts or hoodies with cartoon patterns on the chest, and unstyled hair. Many carried backpacks, like attending a class reunion.

They are post-2000 AI practitioners, most prefer using anime or cartoon characters as avatars, habitually use emoji packs and exclamation points, and during breaks, order a series of iced Americanos and a marshmallow hot cocoa. When seeing a post-2000 AI entrepreneur, an investor reminded us that the best way to quickly break the ice is to bring two cups of milk tea.

Just graduated with a bachelor’s degree, top tech companies or investment firms are offering them high annual salaries of 2 million, 5 million, or 1 million USD. However, when talking about hundreds of thousands of dollars in salary, the young researcher sipping marshmallow hot cocoa speaks casually, as if discussing last semester’s class schedule.

“Eh, I don’t really care, a million or two more or less doesn’t matter,” said another young researcher with a similar view, “Anyway, I want to start a business, I won’t earn much in a few years anyway.” Some big companies and investment firms also want to recruit him.

The large model industry has mass-produced hundreds or thousands of young elites earning millions annually. Big companies have broken past age, rank, and experience restrictions, recruiting young talent with high salaries.

Several headhunters and HR say that top graduates from prestigious universities, interns in core teams of large models, with publications in top journals, and selected for major talent programs at big companies, typically earn over 1.5 million yuan annually. One involved in seed-stage recruitment said, the 2024 top seed campus recruits earn about 1.5 million, rising to 3-5 million in 2025, and by 2026, core positions can pay up to 6 million, with some earning even more.

Before graduation, these talents are already pre-locked in—starting at 2,000 yuan, with the highest exceeding 5,500 yuan daily. Some have received internship offers from Meta, with monthly salaries of $20k, including housing and meals. While most industries consider a daily internship pay of 200 yuan to be good, these figures defy common sense, often prompting people to double-check: “Is this daily or monthly pay?” “Renminbi or US dollars?”

According to industry statistics, in the first quarter of 2026, the average monthly salary of Beijing food delivery workers is just over 10,000 yuan, meaning an intern earning 5,500 yuan a day is equivalent to 10 delivery workers. A researcher earning 3 million yuan a year working for a year equals the diligent work of a 2025 graduate for 39 years. The latter still has to worry about unemployment.

Even at high-paying internet giants, to reach an annual salary of 3 million yuan, you need a master’s degree, 8 to 12 years of continuous work, at least three promotions, each ranking in the top 30%, and be involved in core business or catch rapid growth, to succeed in reaching ByteDance 3-2, Alibaba P9, Tencent T11 or above before age 40. By then, you’re managing dozens of people and have some industry fame.

Today, a 22-year-old recent bachelor’s graduate AI researcher, with no team leadership, no business decisions, and no performance cycle experience, earns the same income.

And they are still very young. “Born in 1998, already considered ‘mid-level’ in the base model team,” said a big company’s intern in a large model team, somewhat frustrated, as he’s older than most interns at the company. Once working late into the night, he looked up and saw a much younger intern, “still bouncing around while coding,” he said with emphasis, repeating, “literally bouncing.”

The relatively young are only 17—big companies are increasingly age-blind, making it hard to say who is the youngest. She was contacted proactively by an HR from a major company, balancing preparing for her high school finals while interning in a large model team, celebrating Children’s Day at her workstation.

“Doing a PhD is a waste of time!” a young researcher wanted to advise friends from Tsinghua’s “Yao Class,” “Why learn so slowly with the smartest brains?” He gave an example: a 19-year-old Stanford student left school before sophomore year, quickly raising $4.5 million for his AI startup, with “at worst, just returning to Stanford.”

One recruiter once advised three or four PhDs to give up their degrees and switch to full-time roles, offering attractive titles and salaries. “If studying is to find a good job, you already have one. And two years from now, it might not be there anymore.” So, several young people chose to drop out midway and leap onto the AI train.

A more radical idea comes from an AI startup founder who plans to let a high school sophomore drop out, work and learn simultaneously, “He can’t learn more at school than here.”

After surveying over 3,500 unicorn founders, global investment firm Antler found that in 2024, the average age of AI unicorn founders is 29, down from mostly 40 in 2020. This number is likely to keep falling—AI enables these smart young people to find higher-paying jobs or even push their net worth into hundreds of millions of dollars.

AI Native: Youth is More Valuable Than Experience

Senior executives at internet companies are a tiny elite managing hundreds of technical staff or achieving breakthroughs in certain areas, with high ranks, good reputation, and stable positions, avoiding the “35-year-old curse.”

Now, past experience is losing its value. A former ByteDance seed team member said that when they first invested heavily in large models, ByteDance continued the tradition of assigning accomplished “big brothers” to new ventures. Several leaders changed, each bringing their core researchers, but results were disappointing. Later, younger Zhou Chang joined and quickly advanced multimodal capabilities.

“This made us realize our previous hiring strategies were wrong,” he said.

If resources are compared, DeepSeek isn’t competing with billion-yuan giants. It has fewer than one-tenth the staff, works half as many hours per person, has never received outside investment, yet entered the top tier of global large models earlier than China’s leading internet companies. We analyzed 84 publicly available resumes of researchers involved in DeepSeek’s three generations of models, over 70% under 30.

One conclusion from ByteDance’s research on OpenAI, Anthropic, and DeepSeek is that in AI, the key researchers truly determine a project’s progress; past management experience and business achievements are less important. “Big brothers” may not lead research teams poorly, but they won’t necessarily make better decisions than researchers themselves. It’s better to let smart young technical talents lead.

An insider said ByteDance assembled its seed team by transferring a TikTok growth product lead to oversee seed recruitment. The hiring logic focuses solely on investment amount and output, not past titles or salaries. They look at how much they pay and how much benefit they bring, with no regard for previous ranks. “Previously, level 3-1 employees were mostly master’s graduates with about five years of experience; now, campus recruits can get the same or higher ranks.”

The new rule is: the more AI Native you are, the more opportunities you have.

Several researchers try to explain this popular concept in big company job postings, funding proposals, and founder speeches. One said, “Their thinking aligns perfectly with large models’ input-output, asking AI first when encountering problems, knowing what to ask next.” Another compared it to “Why do elderly people need to learn to use smartphones, but children don’t? Because children understand what happens when they tap the screen. Same with large models.”

An AI-focused investor summarized more simply: “The younger, the better.” They are all post-2000.

Since 2022, when OpenAI made large language models mainstream, models have gained multimodal, deep reasoning, and coding abilities. The industry sees new tech roughly every few years.

But four years is enough. Someone who chose to study “Computer Vision” during their PhD before graduation, and who hasn’t finished yet, finds the world has changed. If they switch directions too slowly, they risk becoming “the previous generation” of AI.

The longer they work, the more passive they become. An HR from a large model company said they’re hiring AIGC text-to-image, video generation roles in 2024, and tend to look for those with visual algorithm experience. They quickly find that candidates tend to stick to familiar tech, solving problems with proven methods, and only switch when results improve significantly—fresh graduates and “AI Native” talents don’t copy old work, and after switching, performance can multiply several times.

“People with five or six years of experience might switch quickly, but why gamble? There are always younger ones,” said a headhunter after rejecting dozens of resumes. “33 is probably the upper age limit.”

Headhunters have some screening tricks. If a candidate asks about company revenue—immediately considered not AI Native enough. Most AI companies aren’t profitable yet; they care more about compute power, models, and data. Revenue is a financial metric for older companies.

“Genius” managers only want to hire peers. Would a 30-year-old tech lead want to hire someone older or less skilled? questioned a headhunter who worked with ByteDance.

She quickly listed examples: Zhou Chang, who built ByteDance’s multimodal capabilities, is in his 30s; Yang Zhilin founded Kimi at 30; Alibaba’s former head of large models Lin Junyang was born in 1993; Xiaomi’s MiMo large model head Luo Fuli was born in 1995; Tencent’s Yu Shunyu, head of the Mysterious Language Model Department, was born in 1998.

Moreover, most young people are more willing to work overtime. A 21-year-old AI intern often works from 11 pm to past 1 am, taking breaks to eat and walk around, keeping alert, and working “a bit on weekends too.” “It’s not the company’s requirement, it’s my own standard,” he said. “Otherwise, it’s hard to stand out among peers.” Another 22-year-old researcher doesn’t see this as unusual; sometimes he pulls all-nighters from 9 pm to noon the next day because he’s “more immersed.” They are still far from family responsibilities.

High school, free shipping, and finding even younger talent

Large model companies have achieved results with young people, and this awareness quickly spread—companies must become more youthful to AI-ify. Besides AI researchers, product, design, PR, and HR also need more young talent.

Li Xiang, founder of Ideal Auto, announced that 2026 is the last window for sprinting to become an AI leader, and he said on social media that without deep training and learning, most ten-year veterans perform significantly worse than those with just one year of work, and the gap with top 90th percentile campus recruits is at least tenfold—like “gold ore, no need to open blind boxes to extract gold.”

In March this year, Geely Holding Group and ChipWit Technology announced a targeted talent program for high school students to prepare talent for Geely Intelligent and other businesses.

Hiring young people isn’t just about telling a company transformation story; there are practical work needs. An AI-driven payment company said they only consider post-1998 born candidates for media roles because active tech KOLs are getting younger, and they need to communicate with similarly young people. In venture capital, young investors are better at talking with entrepreneurs.

Ultimately, pressure is mounting on senior internet industry leaders. Today’s recognized AI organizations are expected to be flat and transparent. Young talent dislike traditional hierarchical management and prefer meritocracy.

In June, Alibaba quickly replaced DingTalk’s former president Wu Zhao, who had been hired over a year ago, with Chen Yusong, born in 1992. Wu Zhao’s former partner said Wu Zhao is still the same, eager to achieve big things, but “he knows the times have changed, but maybe not how people and society have changed.”

Everyone wants young people, but the real challenge is that truly smart young people are limited. Catching and securing these talents before graduation is crucial. Several HR at big companies said they found that if a “little genius” interned at a big company and had a good experience, the probability they would choose that company after graduation is very high—“smart people are limited, it’s about establishing connections early.”

In Denver, USA, during CVPR (IEEE/CVF Conference on Computer Vision and Pattern Recognition), one of the “top three” conferences, Nvidia, ByteDance Seed, and Intel hosted a dinner inviting young scholars; the next day, Tencent Qingyun, Alibaba Star, and MiniMax held their own. Half a month later, in Seoul, Korea, at ICML (International Conference on Machine Learning), Alibaba, Kuaishou, and Tencent again held dinners on the same day.

Tencent announced that at one event this year, at least 12 leaders would attend. Kuaishou booked a cruise on the Han River, with customized fireworks, and top business leaders from Kuaishou interacted directly with attendees. Alibaba’s dinner was held on the 38th floor of the Grand Hyatt, where Buffett once spoke.

To show sincerity, some companies arrange for department heads, vice presidents, and key interns to connect, have coffee, and exchange views on technology and industry, or discuss life goals. Missing this time isn’t a problem; some HR ask about recent updates, send small gifts for Mid-Autumn or Spring Festival, and suggest considering them when they start working—“the top salary others get is just our starting point.”

A seed-stage person said that around 2026, Seed established a “Student Affairs Department” to screen and lock in interns and fresh graduates. Their database nearly exhausts China’s top students, with lists of students from key universities, labs, mentors, competition experience, and internships.

In theory, if you’re an outstanding student from a top high school, Seed’s HR might know more about where you study, when you graduate, and where you intern than your relatives.

For high-level competitions, they can sponsor GPUs, tokens, or other needed resources for coaches. Besides the list of winners, they also understand each participant’s performance. For example, a low total score doesn’t mean a participant isn’t talented; maybe one of three judges gave a very low score—“a semi-open secret.” One HR said, “Ask around, other companies know too.”

For competitors, big company HR are asked to label relevant teams as much as possible, including daily performance, output, contribution, and technical strengths, asking many people for verification, and matching these with their own team’s needs. If a mentor’s student interned very well, their team gets special attention. Mentors are generally happy to cooperate with big companies, and some students joke about “packing into factories” with classmates.

A student intern contacted by several big companies said that when choosing an internship, the first thing is the team’s reputation—whether it’s a multi-modal, pretraining, or post-training group, and whether they’re willing to do “dirty work.” Second is the workload—without enough tasks, it’s hard to do anything. Third is the team atmosphere—opportunities to directly communicate with experts. Fourth is pay.

Big companies aren’t short of money. ByteDance set up a Top Seed Talent Program for Seed interns, with an average daily internship pay of 2,000 yuan last year. This year, the Top Seed program was officially canceled, but the top salary has no cap. Tencent’s Qingyun plan covers the entire group, with the most spots in the Metaverse and large model teams, paying from over 20k to around 80k yuan per month, with some earning about 110k yuan—another form of competition. Daily pay “day by day,” but monthly salaries include paid leave.

Interns spread sayings like “Seed chooses Seed,” “Goose (Tencent) picks Goose.” If not suitable, there are also “stars”: Meituan’s “Beidou Plan,” Alibaba’s “Alibaba Star,” Kuaishou’s “Kua Star,” Xiaohongshu’s “REDstar.”

Job postings are increasingly earnest, emphasizing not just salary but also what the company can offer researchers, such as “leading core projects,” “unlimited pay,” “join now to take on key responsibilities earlier.” To boost appeal in the talent war, startups like Kimi announced early stock options for top talent interns—Zhipu’s stock price soared 20 times in less than half a year, making those options highly valuable.

After joining, these young talents also gain much more freedom than ordinary graduates.

Some campus recruits admitted through top talent programs are directly managed by business leaders, with some autonomy to decide what’s worth doing, to initiate new projects, report, and build teams around new directions, rather than just optimize existing businesses by 1% or 0.1%. Yao Shunyu often invites Tencent Mysterious Language Model interns for meals and regular exchanges. One intern said he felt “the company hopes to cultivate long-term, and expects you to make a mark at Tencent.”

Some companies allow candidates to team up with peers also selected through talent programs, forming small groups to explore new directions. After joining, some students find their computing power insufficient and include their needs in weekly reports sent to top management. Within three days, their department secures over ten million yuan in computing resources.

The利益链 behind “youth”

In the investment world, “post-2000s” have become a key label for projects.

A 27-year-old researcher, who doesn’t consider himself young, just started a business. To gain market share, an investment firm sent a blank investment intent letter, meaning “conditions are open”—who knows, maybe the “next OpenAI, Anthropic, or DeepSeek” is among the young people carrying backpacks today, which sounds more imaginative than a 40-year-old startup.

“Finally, we’re enjoying the dividends of the era,” said a 2003-born AI entrepreneur who completed two years of grad courses in half a year, dedicating the rest to startups. His first round of funding was tens of millions, with partners two or three years older, and a team of over twenty, many interns from Tsinghua’s AI community—home to many similar startups.

“It’s not much,” said a doctoral senior from the same school, who raised hundreds of millions in a few months. Some classmates launched four rounds of funding within a month, “doubling valuation in place,” he asked, “Do you know what ‘in place’ means?”

“It means nothing has changed. The only difference is the amount on the business plan,” he said. Many investors still come knocking.

After a company’s post-2000 founders signed a financing deal, one of the co-founders resigned in frustration, saying, “This is kids’ entrepreneurship.” But what if that company succeeds? Who would care if Zuckerberg shows up in pajamas and T-shirts to meet investors?

Sequoia’s youngest partner, Cao Xi, who invested in DeepSeek after launching a new fund last year, said that now is the era of 90s-born founders. Six months later, the entrepreneurs he met were born between 2000 and 2002. “Sometimes I think, I wish I weren’t an 80s kid,” he said.

Similar to Qiji Chuangtan, which focuses on early-stage funding for young people, some investment firms are forming funds dedicated to young founders. For example, Yunqi Capital’s Y Transformer, which invests only in founders born after 1998, with a budget of 3M yuan, plans to fund about 20-25 projects, only first rounds, about $600k per deal, decision cycle of 2-3 weeks.

The unspoken industry rule has been “the old boys’ club”: mature tech elites, successful entrepreneurs, and billion-capital investors support each other, “big brothers helping big brothers,” with opportunities, trust, and capital circulating among a small circle. Most core projects are in the hands of “the previous generation” investors; young people don’t know key entrepreneurs and lack decision-making power. A post-2000 investor said he must adapt to “big brothers’ rules,” being clever at toasts and reading the room at dinners, seeking mentorship.

AI has given young investors opportunities—veteran investors don’t understand much, most entrepreneurs are young, so “big brothers” are willing to listen to their younger managers. One veteran said he plans to promote interns, “Just like many AI companies’ ceilings depend on interns’ talent and effort, the future of investment firms might also be decided by interns.”

Mutual help isn’t just among young investors and entrepreneurs. AI researchers have high salaries, high mobility, and companies are eager to hire, adopting “defensive hiring”: even without open positions, they don’t want competitors to poach talent, and are generous in offers. Besides high standards, the talent pool is small, making headhunting a lucrative business.

They hunt and advise clever young people. A headhunter receives a request: if they can find research candidates for three specified teams, they get a 10,000 yuan reward per person, regardless of success. Another company offers 30% of headhunter fees for targeted researcher candidates—only CEOs get such offers elsewhere. “If someone’s salary is $1 million, the bonus is at least 2 million RMB,” said a headhunter.

AI talent is getting younger; besides being more AI Native and “useful,” everyone benefits from youth. A bigger theme is that young people tend to band together, build discourse power, and “fight back” against older generations.

Young researchers produce results, prove their abilities, join big companies or start their own, gaining management authority; they trust peers or even younger people. Motivated to explore, they seek to prove themselves to management or attract young investors; successful investments lead to faster promotions.

“Of course, peers communicate better!” said a researcher who spent time in the San Francisco Bay Area, where 20-year-old founders hire 18-year-olds, and 19-year-old investors select them. They didn’t know each other before, just emailed: “I’m very interested in your paper, my idea is xxx, let’s chat?”

He said some domestic investors still follow “that old way”: exchanging business cards, with a big photo on the left and title on the right. Young people rarely do that; “We don’t have titles.” As long as the ideas are interesting, they’re happy to meet new friends via email. Soon, “I know some friends like you, and we get along,” forming a network. Creative ideas spread like wildfire, and a few smart young people can start a startup, get funding, and compete with resource-rich giants.

No one can stay forever young

In the extreme atmosphere of idolizing youth, a former “Huawei genius boy” faced a comprehensive shock. He graduated with a PhD, earning a salary far above peers, even at top universities, and was envied as a promising hire. Two or three years later, juniors’ salaries surpassed his expectations—ByteDance started recruiting foundational model researchers with no salary cap, often doubling previous offers.

A year later, he started a business, with Tencent and Alibaba also joining the talent war. The salary expectations for top campus recruits are “shockingly high.” He could only use emotional appeals, claiming to be more reliable, offering more options, and recruiting from his alma mater. For fundraising, the “Huawei genius boy” label still worked, but it was less attractive than the rising stars of the post-2000 generation.

Young people come in waves; there’s no “most young,” only “more young.” Competition has become fiercer. An AI industry insider observed that the number of top conference submissions has skyrocketed from a few thousand around 2020 to over 70,000 now. A master’s student used to publish two top conference papers as a good standard; now, that standard has doubled or tripled.

One AI researcher posted interview experiences for top-tier talent programs, creating a chat group requiring relevant internship experience to join. The 500-member group filled up in two days. They discuss interview tips, internal conditions, and many HRs from big companies follow his account “Random Field” to gather info on interns and graduates.

The unwritten rule is: to get into top talent programs, you need a good internship. To get a good internship, you need a good first internship. “How to get that first good internship? Rely on senior students or mentors for strong internal referrals,” said a serious-faced post-2000 intern. “No referrals? Then just gamble.”

Another candidate who got into a top talent program said, “The circle for base models is already closed,” with interns from major model companies moving back and forth, then recommending juniors after becoming full-time, “People inside don’t come out, and outsiders can’t get in.”

“Many facts are better left unknown; revealing them is cruel,” hesitated a recruiter familiar with AI hiring. “In the past, ordinary university students earned 100k yuan a year, Tsinghua and Peking University grads earned 1 million. Everyone accepted a tenfold gap. But now, Tsinghua grads can earn 5 million a year, while ordinary students can’t even reach 50k. The gap is 100 times— isn’t that cruel?”

A post-2000 AI researcher said he feels lucky: “This era’s rewards for extraordinary talent have never been so generous”—but people often focus on the first half of that sentence, ignoring the second: “The punishment for mediocrity has never been so severe.”

That “Huawei genius boy” can at least start a business. Most of his peers studied from undergrad to PhD, went through at least five rounds of interviews, beat out others, and joined big internet companies around 2020, becoming high-paid elites. The age anxiety of “35” exists, but they keep pushing themselves to improve, hoping to outrun the 10% of colleagues who are being phased out.

AI is here. Front-end developers are becoming “redundant,” and other software engineers are only a matter of time—most big company programmers are anxious, trying harder to distill themselves, aiming to outperform colleagues, only to be eventually replaced by AI.

Before the second half of 2025, a senior engineer over 30 at a big company never doubted his “age.” He studied in the US, entered a big company smoothly, kept up with new tech, but suddenly felt that the rapid updates of large models and information were like an unstoppable faucet, turning his past experience into “liabilities.”

Anxiety hit him hard: “Before, I couldn’t read 200 papers a day alone; now, I can work with AI to read 300, 500, or even 1,000.” The question is, “What if I don’t see them?” Every night before sleep, he sets tasks for AI to try to ease his unease.

Hearing this, a post-2000 AI researcher asked in confusion, “What else? It’s like cars replacing horses—advanced productivity will replace outdated ones.”

Hours later, another researcher he didn’t know used the same analogy: “Why didn’t they switch earlier?”

“But the horse-drawn carriage drivers might find it hard to learn to drive cars.” “But society progresses this way,” he replied. “Four words: narrow vision.”

The 30-something programmer listened silently after hearing the story. After a long pause, he said, “We all know no one can stop technology; plugging ears and ringing bells is foolish, just follow along. But it’s hard to explain to them that change isn’t so easy.” He left the big company, seeking new ways to explore technology.

A few days later, he messaged that he again felt the brutal confidence of youth. Chen Yusong, born in 1992, succeeded Wu Zhao, born in 1999, as CEO of DingTalk—this has many complex facets, but his young colleagues summarized it as “taking down ‘Old Deng,’ replacing him with youth, and everything will be better.” He didn’t seem to be in that joyful new world.