Attorney Shao Shiwei | Entrepreneurs in the AI Era Are Being Redefined — Observations from the Xiaohongshu Hackathon Grand Final

First, a data point that left a deep impression on me. In the Claude Code hackathon held by Anthropic this year, there were 13,000 applicants, and in the end only one of the top five had a design engineering background; the rest were practitioners from law, medicine, infrastructure and other industries.

This would have been unimaginable five years ago. Five years ago, if you wanted to build a software product, the first thing you had to do was find a technical co-founder, because you could not get past the coding hurdle. Now, AI has helped you get more than halfway past that hurdle.

What does this mean? It meansthat for the first time, the entry ticket to product innovation is no longer determined by whether you can write code, but by "whether you understand what users truly need."

And this is precisely where cross-disciplinary players have the home-field advantage. At the finals of this Xiaohongshu hackathon, I also saw fellow lawyers who had come to compete on site, which surprised and delighted me.

A lawyer knows better than an engineer what clients and their family members need; a doctor knows better than an engineer what patients are most anxious about at follow-up visits; a person who has used a wheelchair for years knows better than any able-bodied product manager what a person with disabilities truly needs. In the past, these insights were all blocked by the threshold of "I have an idea, but I don't understand programming." Now that threshold is being greatly lowered.

The influx of cross-disciplinary players is the true wellspring of product innovation.

What can 48 hours force out? It forces out the instinct to "start from the pain point."

The format of this Xiaohongshu competition is itself very interesting—a 48-hour closed development sprint. When time is compressed to the limit, people have no time to show off their skills. You can only pick the problem you are most familiar with, that hurts the most, and that you most want to solve.

The most moving projects on site all followed this pattern.

Brain-controlled wheelchairIt won first prize in the hardware track. The developer himself is a wheelchair user who has experienced two episodes of severe paralysis. The device controls the wheelchair through a hybrid signal of EEG and EMG, and he shaped the head-mounted part into the form of a golden cudgel. A person trapped in his own body built himself a golden cudgel with his own hands. This is a product driven by pain.

The device uses a hybrid control technology combining "EEG and EMG." The user wears a headband, focuses their intention, and combines this with subtle movements of the head muscles to control the wheelchair's forward movement, backward movement, and turning. To keep the operating threshold sufficiently low, he greatly simplified the complex control logic, retaining only the most essential few actions.

His wife "Xianggu" is the team's "idea generator." She was originally a web fiction writer and, for this project, forcibly crossed over into becoming a geek developer. Two people, one wheelchair, one golden cudgel, 48 hours. Feiniu said on stage that what he wants to build is not just a product, but also to offer a new possibility to those trapped in their bodies—enabling them to use the faintest signals to re-establish a connection with the world and support themselves in moving step by step toward recovery.

Sitting in the audience, after hearing these words, I realized thatbehind every one of the most moving products in this competition stands a developer who is "a user themselves."

In addition, there were some other interesting products.

Mind KitIt solves the problem of one's train of thought constantly being interrupted during in-depth conversations with AI—splitting the conversation into independent thought nodes, enabling the AI to process multi-line tasks in parallel, and ultimately consolidating them into a thought topology map. The person who built it had had enough of this themselves, so what they built strikes right at the pain point.

There was also a team of junior high school students with an average age of 13, who built a tool that uses AI to help creators produce hit Xiaohongshu posts. In their daily lives they themselves scroll Xiaohongshu and themselves think about how to gain followers, so their understanding of this scenario is deeper than that of the vast majority of adult practitioners.

Judging from the feedback of the audience on site, the popular products did notstart from "what AI can do"; they all started from "what I (or the people around me) am currently being troubled by."

The 48-hour competition format merely magnified this pattern. It forced developers to skip all the "sounds cool" pseudo-needs and strive to find "real problems."

Xiaohongshu deserves separate discussion. What makes it special is not its large traffic volume. What makes it special is that it compresses five things that were previously scattered across GitHub, Twitter, search engines, and e-commerce platforms—discovering pain points, public release, real-time feedback, precise search, and closed-loop conversion—into a single field.

The past path for building products was: have an idea → find money → recruit people → develop → launch → find users. Every step had a high threshold, and every step required capital as a foundation.

The current path is becoming: have an idea → AI-assisted development → post on Xiaohongshu → collect feedback → then decide whether to scale up.

The steps cut out in between are the most expensive steps of the past decade. This means the cost of trial and error in entrepreneurship has been compressed to a level that ordinary people can bear. The price of failing once has changed from "losing all your savings" to "one weekend."

This is what is truly attractive about the AI era—it is not about enabling a few geniuses to make more impressive things; it is about enabling most ordinary peopleto have the qualification to enter the arena for the first time。

After watching these two days, if I were to give a few specific suggestions to entrepreneurs who are on the road or preparing to set out, I would say these things.

First, stop asking "what can AI do" and ask "what has recently annoyed me." All products that start from technology and work backward to application scenarios will lose to products that start from real pain points and work backward to technical solutions. Your advantage is not AI—everyone has AI—your advantage is that you understand the specific pain in a specific scenario better than others.

Second, a cross-disciplinary background is not a disadvantage; it is an asset. If you are a lawyer, doctor, teacher, nurse, renovation worker, or used-car dealer, do not feel that "not understanding technology" is a problem. You have something that people with purely technical backgrounds lack: "industry intuition." This is something AI cannot learn, and engineers cannot learn it either; it can only come from the years you have spent in that industry.

Third, ship it first, then scale it. Do not take the old path of "working behind closed doors for a year before launching." You can have an idea today, build it tomorrow, and ship it the day after tomorrow to see the feedback. Let the market tell you whether this thing is worth building, rather than letting your own self-satisfaction tell you.

Fourth, warmth is the last moat. The products that moved people most on site all share one thing in common: they made people feel that "the person who made this genuinely wanted to help another person." This is true of the brain-controlled wheelchair, true of the middle-school student's note-taking tool, and true of every project that made the judges want to take a second look. AI can help you write code, help you make slides, and help you generate content, but AI cannot generate empathy on your behalf.When everyone uses the same set of tools, the depth of your understanding of people is your only differentiator.

Fifth, execution matters more than anything. None of the people who got results in this competition were the ones with "the flashiest ideas"; they were all the ones who "actually got something working within 48 hours." The biggest illusion of the AI era is that, because the barrier to entry is lower, everyone feels they "can start at any time." But the fact is:Things you can start at any time are, for the vast majority of people, things they will never start in their entire lives.

As a lawyer who has long represented clients in cases in the new economy and Web3 sectors, the reason I came specifically to Zhangjiang to watch the competition over these past two days is also because the client group I serve—entrepreneurs in the AI and Web3 fields—are precisely the protagonists of this transformation.

What I have seen over these two days is a group of extremely intelligent, extremely passionate people who, for the first time, have truly mastered the tools of creation. Many of them will, in the coming years, build things that today's version of us cannot imagine.

But from the specific cases I have handled one by one, I have also seen the other side of the coin:The risks on the entrepreneurial path rarely descend suddenly at a single point in time; they tend to grow quietly out of each specific decision made every day. The design of a product feature, the wording of a piece of promotional copy, an adjustment to a business model, the choice of a business partner—each looks very small on its own, so small that in the moment one would not stop to think about it at all. But one day they will converge into a situation you never anticipated.

The more cases I have handled over these years, the more I feel that a lawyer should not merely be "the person who helps you clean up after something goes wrong." In the AI era, which iterates on a weekly basis, what entrepreneurs truly need is someone who walks beside them and can remind them before they have even seen the bend in the road—someone who accompanies them along a longer path, rather than only waiting for them at the finish line.