During a visit to a newly opened office in Beijing's Zhongguancun district, we met Lu Che (pseudonym) just as a discussion on model experiments wrapped up. Among the roughly twenty people present were Olympiad gold medalists, provincial top scorers in college entrance exams, and an intern born in 2006 who hasn't even finished their undergraduate degree. No one waited for a leader to make a decision, because this company simply has no departments, no OKRs, and even CEO and CTO titles are deliberately omitted.

Lu Che, 22, just graduated from a top domestic university and serves as a co-founder of this foundation-model startup. He published papers at top AI conferences as a freshman and won a best paper award at a premier conference as a sophomore. By conventional logic, this student-led team should most need management training. But Lu Che's answer is almost disarmingly simple: the company's organizational structure is summed up in one English word - "No wall," meaning no departmental barriers. There are no divisions, no clear job assignments, and no hierarchies.

On the day of our visit, their new office had just finished renovation. In a spacious work area, Lu Che didn't rush to showcase their technical prowess. Instead, he pulled out a chair and proudly told me, "This is the ergonomic chair we specially purchased for our team members. Lie down on it - it's really comfortable, and it's a domestic brand." This brings to mind Luo Yonghao's previous entrepreneurial venture in 2012, when he spent heavily on ten imported premium ergonomic chairs, lamenting on Weibo that his tech team colleagues felt "no more back pain, no more leg fatigue" and could seemingly work 24 hours a day.

Over the past six months, as we visited multiple AI and embodied intelligence companies, we initially intended to ask about technology roadmaps, business models, and computing power budgets. Yet an unexpected topic surfaced organically in nearly every meeting room: how to lead teams, how to retain talent, and what kind of environment employees should have for their 10-plus-hour days. These founders are typically under 35, with core teams averaging around 28 years old. Outsiders view them through a "prodigy" lens while also harboring doubts about whether they can run companies effectively. After our visits, the answer is counterintuitive: when it comes to treating people, these youngest founders may be thinking far ahead of many mature companies.

As the post-2000 generation begins producing founders in bulk, Lu Che represents an extreme sample. From day one, he rejected hierarchical organization. His company has only a few dozen people - half self-described "little geniuses" and half engineers poached from big tech firms to turn wild ideas into reality. Why no OKRs? Lu Che explains that true core innovation often initially emerges between just two or three people - "a top scorer plus a gold medalist collides to create something, then we pour resources into scaling it up." In his view, if innovation requires departments and metrics to drive it forward, it's likely already dead. He often compares the team to a speedboat: "Although small, every person on board is a helmsman who can decide where the boat goes." Strategy meetings are open to everyone; if a 20-year-old intern proposes an idea that proves feasible in small-scale experiments, it could change the entire direction of the main model's architecture. Internal arguments are routine, but he emphasizes, "We argue not because we look down on anyone - technology should spark in friction. Every time we argue, we get excited because we know something new is coming." Who ultimately determines direction? Not any single person. "If the direction of collective effort is decided by one person, it will inevitably be locked into that person's path dependency." Clearly, despite his youth, Lu Che is more wary than most of authority suppressing innovation. Technology is egalitarian - "everyone has the right to innovate."

While Lu Che "dissolves power conceptually," Jiang Yi (pseudonym), founder of a humanoid robot startup, took the opposite path - a detour - before arriving at a similar place. His company gained fame last year through a robot competition. Jiang Yi, in his twenties, founded the company in 2023, calling himself part of a "grassroots team, not starting with halos." At the toughest moments, he said, "we had to fight an extremely fierce battle with limited ammunition and limited rations." He reviewed the biggest mistake he made since founding: early on, he was afraid to be the decision-maker. "I used to gather everyone for discussion, and if my opinion differed from the group's, I'd go along with them." He paused before revealing the psychology behind this: "At that time, my real thought was: if you made the decision and something goes wrong, it's not on me." He admitted, "I was very incongruent and internally drained back then. I felt my thinking was right, but if someone suggested a different direction, I'd say, okay, your call."

The turning point came after a viral success, when the company tripled in size and his confidence solidified. One startup dismantled hierarchy from the start; the other learned accountability through failure. Their paths diverged but share the same foundation: an instinctive wariness of "using position to pressure people." At Jiang Yi's company, during the most intense crunch, he personally stayed up until 4 AM fixing code. At Lu Che's company, you won't find a private office because there's no such thing as "someone's office" to begin with. Neither has read many management classics, yet both answered the same question in their own ways: in an industry driven by brainpower, creativity, and rapid iteration, the hierarchical "top-down command" model may be fundamentally counterproductive.

According to the Hurun Research Institute's U25 China Entrepreneur Pioneers List released in July 2026, 44 individuals under 25 made the list - 19 more than the previous year - including two who have already built robot unicorns valued at over 10 billion yuan, with the youngest honoree just 19 years old. In Silicon Valley, the same script plays out more aggressively: Michael Truell, born in 2000, built AI coding company Cursor to nearly $10 billion valuation with a team of just 40-50 people; Brendan Foody, born in 2004, dropped out of Harvard and brought AI recruiting firm Mercor to $2 billion by age 21. When "post-2000 CEOs" become an organized cohort, it presents a new proposition about organizational forms.

If flat organizational structures can still be dismissed as imported Silicon Valley management style, our deeper conversations revealed that this new generation of founders' care for employees lands in extremely specific, even granular details. Take the most common investment: one world-model company gives every intern a monthly $2,000 budget for AI coding tools, with no upper limit for full-time employees. This is no small sum. Lu Che crunched the numbers: one person operating four agents can do the work of ten engineers, and over 80% of their internal code is written by coding agents. "In the AI era, TOKEN is productivity," he reasoned. "If you want to maximize someone's innovative energy, don't limit their productivity tools."

Beyond tools, there's recognition that "people get tired" - and this recognition often comes with founders' self-reflection. Before preparing a major annual public project, Jiang Yi's company threw everyone into the effort. He recalled an algorithm engineer who, to monitor training results, set alarms before sleep to wake every two hours, check models, adjust parameters, retrain, then lie back down - several times a night. This made Jiang Yi feel deeply ashamed. It's become the norm at this embodied intelligence startup. Jiang Yi has worked until midnight every day since founding the company, and that hasn't changed. His parents urge him to be grounded over the phone, but he responds that he hasn't even taken a few days off on an island - "still standing guard at the company."

Lu Che told us his four undergraduate years were "like doing an extra PhD" - starting in the lab as a freshman, choosing "free exploration" while others chased GPA. In other words, this generation of founders are themselves veterans of "high-intensity sitting and high-intensity thinking." Their sensitivity to employee well-being isn't cultivated through HR training programs - their own bodies remembered it first. At AdventureX, one of the largest hackathons in China, where tech practitioner concentration peaks, the most sought-after commodity on site became chairs. Appearing there was a young domestic brand called LiberNovo - precisely the type of story these tech founders would buy into. Its founding team previously held key positions at robotics companies like DJI and CloudMinds, with core leaders steeped in the robotics industry for years.

The founder himself faced over a decade of high-intensity sitting, with chronic back pain becoming an occupational hazard. After trying virtually every premium office chair on the market costing over 10,000 yuan, he identified a common flaw: regardless of price, the underlying logic was "static support plus manual adjustment" - the person conforming to a chair's fixed curve rather than the chair adapting to the person. So he assembled a cross-disciplinary team and rebuilt the chair using robotics principles. The first product, the OC1, features a "holographic follow-up system" with 4 sets of linkages and 60 joints, enabling headrest, backrest, seat cushion, and armrests to move in millisecond coordination. The integrated flexible backrest, paired with electric assist pushrods, actively shapes to different spinal curves. Simply put: the user doesn't need to adjust anything - the chair follows their movements.

Li Ming, editor-in-chief of FutureWork, told us that AI and embodied intelligence companies represent the core leading industries in China's tech transformation, facing global technology and commercial competition from day one. To compete at that level, the most critical prerequisite is attracting and retaining top talent. "These founders themselves are a generation that endured long-term R&D and high-intensity work, so they understand viscerally how important a good office environment and healthy work patterns are to people's state and creativity. They treat office experience investment not merely as cost but as investment in people - hoping that better environments and healthier work styles will attract, retain, and maximize the performance of exceptional talent." For these "product-manager-style" founders, choosing a chair is the same decision as choosing a GPU or a development framework. A dynamic ergonomic chair built with robotics thinking and a static one built with furniture thinking are two different species in their eyes - the latter locks users into a fixed angle demanding "proper posture," while the former supports users in every moment of leaning back or turning to see a colleague's screen.

In August, news spread that ByteDance had initiated domestic replacement of office chairs: at core work areas in Beijing's Fangheng and Dazhongsi, and Shanghai's Xinjiangwan, the standard-issue Humanscale World chairs - retailing at over 8,000 yuan and colloquially known as "ByteDance chairs" - were swapped for the LiberNovo S1, a brand founded less than three years ago. The price is only one-third, yet the direction is an upgrade. According to internal ByteDance notices, "many employees reported the original chair's backrest couldn't lock and lacked support." Reports indicate ByteDance had employees trial and vote on chairs, with international brands like Herman Miller among the competitors, but based on final voting results, the LiberNovo S1 defeated international brands and was selected by ByteDance employees.

Viewed on a longer timescale, this chair replacement reads like a metaphor. The world's first ergonomic chair was born in 1976, its forward-tilting design responding to the paper office era's hunching-over-writing posture. Fifty years later, screens and keyboards have reshaped work postures - people sit for 10+ hours, leaning, reclining, constantly shifting centers of gravity - yet most ergonomic chair design logic remains stuck in 1976. A 10,000-person company's collective chair replacement, coinciding with multiple AI unicorns' independent choices, converges in the summer of 2026. They point to the same thing: when work's primary vehicle becomes "human + screen + AI," every detail serving people in the office environment gets re-examined.

Interestingly, both LiberNovo's clientele and LiberNovo itself share a common theme: creation - innovation from zero to one. When code can be written by AI, humans become the only irreplaceable variable. To understand this meticulousness, one must first grasp the brutal talent market these companies face. According to a report from Liepin Big Data Research Institute, algorithm engineers are the scarcest AI roles over the past year, with demand accounting for over 67%, and positions with annual salaries above 500,000 yuan comprising about 30%. Qiancheng Wuyou's top position data shows large-model and AIGC algorithm engineers' average annual salary reaching 650,000 yuan. At the top end, numbers are even more extreme: in March, a leading humanoid robot company offered its chief scientist for embodied intelligence a starting salary of 15 million yuan, up to a maximum of 124 million yuan annually. One headhunter revealed that a major internet company offered a Tsinghua/Peking University algorithm PhD intern a daily rate of 5,000 yuan. In Silicon Valley, Meta's total compensation packages for top AI researchers exceed 100 million dollars.

Calculating this down to a chair makes the logic even clearer: a top engineer costing over one million yuan annually spends 10+ hours daily at their workstation. If a chair can give them one extra productive hour per day, uninterrupted by back and neck pain, this investment has virtually no downside. The deeper driver is the reshuffling of production factors in the AI era. At Lu Che's company, 80% of code is generated by agents; computing power can be rented from cloud providers; open-source models are readily available; even "writing code" - once a core barrier - can be 70-80% handled by AI. When everything can be procured at scale or generated by models, the one thing that can't be replicated is a top-tier person's judgment, intuition, and creativity in a given moment - what one might call "technical taste."

And creativity only emerges when people feel relaxed, trusted, and well-treated. By any measure, it's hard to expect exploration into uncharted territory from someone whose back hurts, whose chair digs in, and who's constantly reminded by hierarchy that they're "just an executor." This is why these young people would rather decline big tech's fences. Lu Che notes that many team members hold offers from major companies worth millions or even tens of millions, yet choose to stay. "More and more people tell me they don't want to go to big tech anymore - it's a kind of deterministic input-output. Dark horses shouldn't be penned into fixed tracks and enclosures." To let dark horses run, you first dismantle the fences.

One of the most active investors backing this generation of founders, Gu Heng (pseudonym), a managing partner at a hard-tech early-stage fund who has reviewed hundreds of young teams and invested in 30-40 robotics and embodied intelligence companies, offers a more distilled framework from the capital side. "Every year-end review of our investments, no matter how we sum it up, always comes back to one word - taste," Gu Heng told us. To "elevate taste," he places decorations and fresh flowers in every room of his office. The top founders he sees are never one-dimensional: a memory chip industry leader who writes excellent calligraphy, or a large-model unicorn founder who once played in a band. "People who truly build great companies are people with taste."

This logic of taste neatly explains the granular details we observed in our visits. Giving interns unlimited AI tool budgets is taste. Replacing employees' lumbar-supporting chairs from 1980s design logic is taste. Not using rank to pressure people, allowing a 2006-born intern to challenge the founder's ideas in architecture meetings - that's taste too. The answer isn't in PowerPoint presentations; it's in the ten hours of daily seating posture. Returning to the original question: can such young people truly run companies well? A former insider at a major tech company once remarked to us that in the current large-model race, domestic giants possess the strongest resources and highest talent density, yet repeatedly get surpassed in model quality by startups - DeepSeek being one example, Kimi another.

We initially arrived with a scrutinizing lens, wanting to see how "prodigies" performed in management courses. Instead, we witnessed something entirely different: no walls, no imposing titles, permission for debate, permission for interns to challenge direction. They provide unlimited productivity tools, treat office chairs as seriously as technical selection, and internalize employees' late-night struggles as their own responsibility rather than PR ammunition. The notion that "young people can't manage companies" looks more like an untested bias - just as many assume the large-model race is resource-intensive and that big tech will inevitably win. Perhaps precisely because they're young, they lack path dependency, haven't yet learned to treat people as "human resources," and thus stay closer to management common sense: treating people as people.

Gu Heng says early-stage investing is about "waiting quietly for flowers to bloom, feeling like a spring breeze." But bloom requires soil that's been tended first. An office without walls, a chair that moves with you, an uncapped tool budget, a debate that allows the youngest to say no - these are the soil. They may never be decisive factors in a company's trajectory, yet they daily determine whether the smartest minds are willing to create their best work here. In 1976, the first ergonomic chair answered the hunched posture of the paper office era. Half a century later, these youngest tech companies use a chair that moves to answer a new question: in the era of humans working alongside AI, what is a person truly worth, and how should they be treated? Perhaps this is the real, unwritten consensus among AI unicorns.