Invest in ByteDance alumni. Invest in people who've been broken and rebuilt. Invest in people with outsized ambition and minimal ego.

@Zhiyan Chen

Who's making the most systematic bets on ByteDance alumni?

Creek Stone, a new fund, might be going the hardest.

Ming Chaoping, founder of YouWare, formerly led tools and community at CapCut.

Pan Yuyang, founder of OdyssLife, was among the earliest product managers at Coze.

Bai Te, founder of AirJelly, owned the Minecontext project.

The founder of Karpo handled model strategy for core Doubao features.

The founding team at Shumei Wanwu brings together three members of the legendary "seven-person team" that created Douyin: CEO Ren Lifeng, who led Douyin and TikTok's early business operations; operations lead Li Tian, formerly Douyin's head of operations; and 3D product lead Zhang Bo, an original product manager for both Douyin and TikTok.

Of the eight companies Creek Stone has backed, only one founder has zero ByteDance DNA.

Like the PayPal Mafia, betting on a ByteDance Mafia — is this the right play? It's too early to tell. Optimists see it as an aggressive talent strategy well-suited to early-stage investing; skeptics call it pandering to top-tier VCs' preference for picking up later rounds.

But for a new VC firm, it's certainly a bold entrance.

Since the so-called VC 2.0 era began, I've tracked the rise, fall, and fragmentation of many new funds. Most who chose to leave, beyond the call of new technologies demanding new investors, carried some version of a "why not us?" conviction. This burns especially bright in Creek Stone's two partners, Zhong Luhuan and Li Yihao.

By industry convention, neither had a dazzling track record before this. They didn't even know where their first LP would come from. But they wanted to do it, so they started.

You gotta show up to play.

This is elsewhere's third new fund story, following Nebulon Ventures and Source Code Rhythm.

Creek Stone is barely a year old. But in them, you can see a new fund pushing differentiation to its limits: how they raise capital, find young founders, help them build, and reimagine the partner relationship.

Sincerity? Bullshit!

From April 2025, when they started Creek Stone, through October — a full six months — the two raised not a single dollar from institutional investors beyond themselves and friends.

Their previous résumés became a burden. The ByteDance strategy was repeatedly dismissed: "You have no advantage here."

Once, Zhong Luhuan met with the head of a market-driven fund-of-funds. The question came: What's your differentiation? Zhong thought and said: Sincerity. The response: Bullshit.

The first few million dollars came from their own pockets plus friends' support. Repeated fundraising failures left them prepared to go all-in, investing whatever they could scrape together, "because we believe in these founders and what they're building."

What broke the logjam was finding a path to raising from other GPs. They realized traditional FoF evaluation frameworks couldn't yet comprehend a "non-standard" species like Creek Stone, so they pivoted to GPs closer to the market, on the front lines — a strategy many new funds and solo GPs employ.

As their "ByteDance strategy" yielded early hits, partners at top investment firms began committing personally or through their institutions. This trust based on professional judgment let Creek Stone close its first fund in early 2026, at tens of millions of dollars.

The final LP list ran to 17 names, an extremely dispersed mix of individuals and institutions. Beyond well-known firms and angel investors, there was ¥500,000 or ¥1 million from a middle school classmate, and money pooled by "fellow old-timers" who felt "our institution can't invest, but I'll personally back you."

The ByteDance Strategy

Li Yihao says the ByteDance strategy was stumbled upon like blind men feeling an elephant.

In 2022, changes to Hong Kong's 18A listing rules sent biotech into a capital winter. Li, who had already scored some investment wins in biopharma, thought he might be out of a job. When ChatGPT emerged, he went on a self-teaching frenzy, immersing himself in papers on Transformers, KV Cache, and attention mechanisms, often calling his twin brother — writing code at Google in Seattle — in the middle of the night.

Then he started meeting people. First to Silicon Valley to connect with researchers, then broadly across China, building his talent database and updating it at 450–500 people per year. "I spent more time with founders and industry experts." The database gathered "people who actually build things" across product, engineering, and technical dimensions.

Through constant dredging, Li discovered something: Who's closest to the models? The answer: ByteDance.

"By 2024, every product person at every model lab had ByteDance on their résumé." Li's read: whether Doubao, Seed, or Flow teams, handling real model feedback daily had given them a understanding of technology that was already leagues ahead.

Even this talent pipeline kept extending forward. One data point still gives Li chills: ByteDance employs 1,500 HR staff internally. "These people work so granularly that they'll scan every top Chinese high schooler competing in IOI and AML contests, sending out internship offers in advance."

To him, this meant "the young people closest to the models, with the strongest problem-solving abilities, are almost all buried in their Feishu docs."

So Creek Stone's logic became simple: find the geniuses with 6-to-12-month market-leading cognition in the densest talent pool.

Loser's Mindset

A WeChat group for Creek Stone and one of its portfolio companies was named "Doraemon" by the founder.

Zhong and Li define themselves as founders' "zero-salary employees," even "kneeling servants." They say Creek Stone's core is a "loser's mindset."

This was also inspired by Su Hua.

They'd brought founders to meet Su several times. They observed that Su would listen patiently, then say "I have an idea" or "I have a perspective" — never telling a young person "this won't work."

In Li's view, "Su Hua is the king of loser's mindset."

Zhong had worked many different jobs and investments, made "several buckets of gold." But when his daughter once asked "Daddy, what do you do?" he realized besides "making money" he didn't seem to know anything. This hunger for purpose drove him into Creek Stone. During the Qingming holiday, to help a portfolio company's AI necklace find distribution, he flew to Japan alone to run sales channels.

This methodology also shapes their preferred founder profile: "Extremely ambitious, extremely low ego."

In Creek Stone's framing, ambition is the engine, ego is drag. AI evolves exponentially. A founder with big ego develops "cognitive arrogance." That arrogance makes them an opaque container, unable to absorb the high-frequency feedback flowing from the outside world.

Currently, close to half of their portfolio companies have co-founders recommended by Creek Stone.

In one case, to find a founder the right co-founder, the two spent three months arranging dinners, mixing and matching combinations. Finally, after one hotpot lunch, the other party resigned that afternoon, dyed their hair red the next day, and moved to Shenzhen. Another time, to bring a seed-stage talent who only wanted to be "number one" into a portfolio company as co-founder, they spent three months matchmaking and observing before a "dream team" formed.

Another project, when the founder was slightly uncertain on product direction, brought in a "co-founder elder" with deep industry expertise — within a month, they'd built a complete "industry-changing product."

"We must respect founders! We must love founders! We must be good to them! This is the bottom line!" Li loves exclamation points. He says, "What convinces the next founder to take Creek Stone's money is often our previous founders."

Scars

As Creek Stone kept investing, a hidden pattern emerged: these founders' lives had all endured some kind of "scar." This parallels London-based early-stage fund Hummingbird Ventures.

One founder they backed, born in excess of family planning limits, didn't learn until age six that his birth mother was someone else. He was then taken to an unfamiliar home, where his biological sister locked him outside and refused him meals.

Another was a top student all the way through, until his grad school recommendation spot was taken by a "second-generation" privileged kid, sending him to a mediocre school. This fall from grace gave him a built-in "rage." During a ByteDance internship, this rebelliousness was spotted by a boss who bypassed all rules to get him onto the core team.

"Being fucked over hard by life,消解痛苦 through massive reading in solitude — this is the only fuel that lets someone survive the dark nights of entrepreneurship."

This empathy for scars also gives Creek Stone a感性色彩.

To some degree this projects traits of Zhong and Li themselves — at their previous firm, both had repeatedly pushed MiniMax, Zhipu, and Moonshot AI through early-stage processes, only to watch them die in large-institution machinery over "we don't get it" or "the numbers don't work." Too scared to invest early, then jumping in at double the valuation once big names had led later rounds — a loop they couldn't escape for a time.

What broke Li most: a $500K early-stage company whose internal compliance demanded the founder sign unlimited personal liability. He smashed his phone on the spot in fury.

Zhong once tried explaining to superiors why post-'95s understood AI better. He got silence or dismissal. When he resigned, his boss publicly lectured him at length on "how to do VC" — a public "execution" that "heretics" in that system must endure, he says.

AI Products Are JDM

AI product bubble narratives are already VC gospel. To these claims, Li — a car guy — compares current AI products to Japanese domestic market performance cars, JDM.

"They're all半成品. Companies and users have to驯化 them in use. Harness isn't worth money, Hermès isn't worth money. What's worth money is performing a 60 first, then selling to users or companies. As they use it, it goes from 60 to 90 — the product becomes theirs."

Creek Stone manually scores GitHub projects to train their own "lobster." "You find that after scoring the second batch, the AI's aesthetic aligns with ours."

They also have this "半成品" nurturing feeling toward founders. Want to start a company but have no direction, no co-founder, no market or customers? Not the biggest problem. The talent database kicks into a new gear.

Zhong and Li bring different but complementary people together, waiting for chemical reactions. In Creek Stone's portfolio, half the co-founders came through them; some teams had their first several hires entirely handled by the firm.

"The most important thing is bringing great people to the founder."

Meanwhile, the atmosphere of entrepreneurship "lures" the next person hiding a restless urge to create. "Starting a company is multi-dimensional. Watching someone similarly excellent suddenly expand across dimensions — some people can't suppress that impulse."

Recently someone asked: What's the North Star metric for AI-native?

Their answer: First, total token consumption. Second, depth of context entanglement.

ENFJ + INFP

Zhong Luhuan is ENFJ, Sagittarius. Li Yihao is INFP, Aries. Li's high school desk mate happened to be Zhong's middle school classmate.

In Zhong's eyes, Li is "terrible at managing up" but naturally built to be a hunter. In Li's eyes, Zhong is the big brother he'd trust with his back.

Their radically complementary personalities shape Creek Stone's decision process: Li mainly figures out "can this work," Zhong judges "can this person make it."

Despite being completely different personality types, their preferences align with eerie consistency across all projects. They even force themselves to evaluate deals separately to prevent aesthetic overfitting — otherwise they meet and discover they've both fallen for the same one again.

Only one project has created major disagreement. Li thought the business was totally fine, investable. Zhong vetoed it: he couldn't feel the founder's "sincerity."

The Office That Doesn't Spend Management Fees

Creek Stone's office sits on Shanghai's Bund, 133 Yuanmingyuan Road, in the former Chinese YWCA building erected in 1932.

The fifth-floor space is half free desks and meeting rooms for a dozen-plus startup teams getting started, plus a hotpot room for impromptu gatherings. Two or three portfolio companies have already outgrown these desks and moved to larger offices.

The other half is designed as a stylish bar and open space. On a low coffee table sits a Gengar Pokémon figure wearing a black baseball cap embroidered with Creek Stone. From this side's windows, you can see tourists posing for photos below, the weekly rotating market. The old office of Zhong and Li's previous firm once stood nearby.

Every LP visit draws the same exclamation: "This is nicer than our office." Zhong always smiles politely: "Didn't spend management fees on it. Our own money."

Stepping Stones

The name Creek Stone came from a four-model bake-off after they decided to quit.

That day, they fed the same prompt into ChatGPT, Gemini, Doubao, and Kimi: a nature-related name, timeless yet resilient.

From each model's dozen-odd suggestions, they chose Creek Stone: Creek, a stream; Stone, a rock. Creek is the source of water's highest goodness; Stone is the backing that grinds forward. Zhong says it's a particularly "them" name: absolutely no desire to be hunters sitting high above, judging the landscape — but rather stepping stones in the stream of the era, the ones who most understand founders.

The model that contributed this name was somewhat unexpected. Not the most advanced-seeming one, yet somehow fated — Doubao.

Cover image: Edvard Munch, The Sick Child, 1885–1886, National Museum of Norway