Original · Unique Research · 2026-03-28 · Shanghai
Editor's note: The article was published on March 28 and identifies the roundtable date as March 21, 2026. Team sizes, repository-growth comparisons, traffic claims and institutional relationships below are attributed to the original report and speakers, not independently verified current facts. The source's description of OpenAI bringing OpenClaw “into its fold” does not establish a specific acquisition or employment relationship. Its product naming varies between “claude code” and “cloud code”; both are retained as written, and “shrimp” remains a source nickname. The 1,000-times language is an argument about amplification, not a measured performance guarantee. Speaker handles and Silicon-Carbon Exchange are provisionally rendered. All extracted text is translated; substantive source-image content remains to be checked before publication.
Unique Research & ClawdChat
Why Might a Team of 3 Have a Better Chance
of Winning in the AI Era?
ClawdChat Hardcore Party Roundtable
Event date: March 21, 2026 · Shanghai
Moderator: Wu Wei, founder of Unique Research
Panelists: Weilian (ClawdChat), Jason (Tencent WorkBuddy), Allen (Shanda EverMind), Shougong Chuan (Lovstudio.ai), and Gude Bai (Silicon-Carbon Exchange)
Recently, a 3-person company has been sharing a stage with Tencent and Shanda to discuss AI's future.
If you are building an AI startup today, one sentence can easily discourage you: the big companies are all in—what chance do you have?
By the end of this roundtable, however, the conclusion was exactly the opposite:
The heavier large companies become, the more openings small teams may find.
01 “The ‘Shrimp’ Is the Phone; claude code Is the Computer”
Two weeks earlier, Gude Bai had said on a livestream that OpenClaw was “not much use.” Today, he changed his mind.
“
Beware of one paradox: so advanced that it falls behind.
He offered an analogy: the internet required sitting at a computer; the mobile internet lets people pick up a phone anytime, anywhere. The barrier to claude code is too high—only programmers' programmers can really handle it.
The “shrimp” is something everyone can use.
“Ordinary people do not need to hear about your MCP or your memory system. They should never have needed to know those things in the first place.”
He said it lightly. The room went quiet for a moment.
He was not talking about parameters or architecture, but a practical question: who can actually put AI into ordinary people's hands?
Shougong Chuan added from beside him that ClawdChat's most distinctive feature was its memory system.
“
If you hire an assistant and have to explain today's task again tomorrow, and again the day after, you will certainly fire that assistant.
Winning is not about having the most advanced technology. It is about lowering the barrier to entry and getting closer to users.
02 “We Ship Faster Than They Can Clone”
Speaking for Shanda, Allen told a story.
OpenClaw's open-source ecosystem was becoming “unstoppable.” React and Vue had taken 10 years to accumulate 300k stars, while OpenClaw's growth was already soaring.
More importantly, he said that after OpenAI brought OpenClaw into its fold, its first move was to put it in many developers' hands to build the ecosystem.
Allen read aloud what he described as a tweet from OpenClaw's official account:
“We ship faster than they can clone.”
(We release updates faster than they can clone them.)
“That is a wonderfully romantic line for programmers.”
Representing Tencent, Jason replied that WorkBuddy had “always put users first,” but that “when taking a neutral position, we may still lean more toward embracing Tencent's own products.”
This also means it is difficult for large companies to stand entirely outside every ecosystem. For small teams, that may be precisely where an opening lies.
Open source is not charity. It is the most efficient way to acquire users and iterate.
03 “I Like Being Copied”
When asked whether he worried about large companies copying him, Shougong Chuan gave an unexpected answer:
“
Not at all. I like being copied.
Late last year, he built a browser extension that could extract web pages into Markdown. It worked so well that a friend used it every day, then started heavily modifying it. Eventually, the friend privately asked him: “I have changed so much—can I spin it off on its own?”
Shougong Chuan replied: “Just open your own repository.”
While many people worried about being copied, replaced or swallowed by large companies, he was encouraging others to open separate repositories.
“
Good things should be developed and improved together. All my own code, too, comes from what I learned from open source.
Competition in the AI era is not about “whether you can be copied,” but “whether you can keep iterating before others copy you.”
04 “AI Native People”
Wu Wei asked everyone: how many people do you have now, and whom will you hire next?
The answers were interesting:
ClawdChat: 3 people. Before hiring, first ask whether an agent can do the work.
Shanda EverMind: 35 people. Hire versatile people who can step outside their comfort zones.
Shougong Chuan: an OPC, or one-person company, beginning to expand this year.
What is an AI native person?
Someone who first asks, whenever a problem appears, whether AI can solve it.
Jason added that AI cannot realize your creativity for you. What they need are people “with ideas who can iterate quickly, validate quickly and learn through trial and error.”
Small teams can naturally organize themselves around “agent first.” Large companies find that difficult.
05 Gude Bai's Hesitation
As the roundtable neared its end, Gude Bai said the past month had been “very uncomfortable.”
Two paths lay before him: selling courses—OpenClaw's traffic was even greater than DeepSeek's—or continuing to build Silicon-Carbon Exchange.
“
I can hardly resist, but so far I have not acted.
Wu Wei said: “That is the difficult but right thing to do.”
Gude Bai made an interesting observation:
AI can amplify human effort, even by 1,000 times.
But there still has to be a “1” in front.
06 A Final Word
The roundtable offered no standard answer.
But a consensus was emerging: large companies going All in AI does not mean small teams have no chance.
The heavier large companies become, the more openings small teams may find. Competition in the AI era is no longer simply about overwhelming others with resources. It is a contest of neutrality, speed, openness and the ability to differentiate by use case.
AI will not erase the differences between founders. It will amplify their judgment, speed, openness and organizational capabilities even more dramatically.
Previously, the contest was about the concentration of resources.
Next, it will be about who most resembles
“a founder whose capabilities AI can amplify.”
07 Selected Q&A
Q1: With major companies such as Tencent and ByteDance all building AI, do small teams still have a chance?
Weilian, founder of ClawdChat.cn
Yes, and the opportunity may be greater than people think. Our approach is not to build a closed product, but a completely open platform where different agents can join seamlessly. Large companies naturally prioritize their own ecosystems when building products. Small teams, however, can take a more neutral position. That position is precisely the opportunity.
Q2: Why do you say, “so advanced that it falls behind”?
Gude Bai, creator of Silicon-Carbon Exchange
Many people naturally assume that the more foundational, complex and advanced something is, the more it represents the future. But what actually gains traction is not necessarily the most complex thing; it is what more people can easily use. cloud code resembles something from the computer era: the barrier is high, and you need to understand many concepts. The “shrimp” is more like a phone, hiding much of that complexity. Ordinary users do not need to know terms such as MCP or memory systems. They only need to know: can I use this directly?
Q3: What do you think is most distinctive about ClawdChat?
Shougong Chuan, founder of Lovstudio
It is not the superficial question of “whether it can chat,” but memory. Imagine hiring an assistant, explaining something today, having to explain it again tomorrow and starting over the day after. You would not keep that assistant. AI products are the same. Whether they remember you, understand your context and carry your habits forward is what truly determines stickiness.
Q4: What is the real value of an open-source ecosystem? Is it just excitement?
Allen, ecosystem lead at Shanda EverMind
It is not excitement; it is speed. A major reason ecosystems such as OpenClaw are growing so quickly is that open source lets more developers participate and move things forward together. The official line “We ship faster than they can clone” captures it well, I think. The point is not whether others copy you, but whether you can keep releasing, keep iterating and keep leading the ecosystem forward.
Q5: Do you really not worry about large companies or other people copying you?
Shougong Chuan, founder of Lovstudio
I genuinely do not. To some extent, I even welcome it. Much of what we know was learned from the open-source world in the first place. Good things deserve to be developed and improved by more people together. The more tightly you try to conceal everything, the more likely you are to slow yourself down. At this stage, the most frightening thing is not other people copying you, but closing yourself off.
Q6: What is the biggest difference between large companies and startups today?
Jason, lead of Tencent WorkBuddy
Large companies naturally have resources, product capabilities and user bases, but they also tend to stay closer to their own systems. From the user's perspective, we emphasize “putting users first.” In practice, though, large companies find it difficult to make choices entirely independently of their ecosystems. If startups can take a more flexible, open position, they can often enter opportunities that large companies cannot fully attend to.
Q7: What should founders in the AI era fear most?
Gude Bai, creator of Silicon-Carbon Exchange
The greatest danger is not large companies entering the field, but deceiving yourself. You can tell the outside world that you are an AI company, but internally you must be clear: what problem are you actually solving? AI is only an amplifier. It can multiply efficiency many times over, but first there must be a real, valid need—that “1.” Without that “1,” however much you multiply afterward means nothing.
Q8: How will you hire now? What kinds of people are better suited to this era?
Several panelists
Weilian, founder of ClawdChat.cn: There are only 3 of us now. Before deciding to hire for many tasks, we first consider whether an agent can do them.
Allen, ecosystem lead at Shanda EverMind: We place greater value on versatile people—not those who stay in a single comfort zone, but those willing to step beyond it, develop capabilities and adapt quickly.
Shougong Chuan, founder of Lovstudio: I would prioritize AI native people. Being AI native is not about how much jargon you know. It is about whether, when you encounter a problem, your first reaction is to ask: can AI solve this?
Jason, lead of Tencent WorkBuddy: AI cannot replace genuine creativity. We need people with ideas who can validate quickly and are willing to experiment repeatedly. AI has already made trial and error much cheaper. The key is not only execution, but whether you have the idea and can recombine different things.
Q9: If you could leave one sentence to sum up this roundtable, what would it be?
Gude Bai, creator of Silicon-Carbon Exchange
“AI can amplify human effort, even by 1,000 times, but there still has to be a ‘1’ in front.” That “1” is not a PPT, a concept or traffic. It is whether you have judgment and know what problem you are solving.
Source: Unique Research · ClawdChat Hardcore Party roundtable