Original · Unique Research · 2025-12-31
Historical edition: This complete textual article was published on December 31, 2025, reporting an event held on December 28. Roles, market-leadership claims, business figures, product capabilities and GEO outcomes are the original report's accounts of speakers' statements, not independently verified current facts. Chinese organization names below are transliterated where no established English name is supplied. The closing January 23, 2026 event notice is preserved as a historical announcement, not a current invitation. Images and image-only data are outside this text-only edition.
On December 28, 2025, “Unique OpenDay: AI Entrepreneurs' 2025,” jointly organized by Unique Research and the Wu Xiaobo Channel, concluded successfully in Xiamen. At a moment when generative AI is moving from wonder to normality, Jason of Unique Research and five AI entrepreneurs working deeply in different fields avoided empty discussion of a grand future and instead shared genuine insights from trend analysis, shipping, marketing, supply chains, cross-border e-commerce, and technical deployment.
From avoiding head-on competition with major companies to finding high-value scenarios, and from delivering tools alone to promising delivered results, the event revealed the pragmatic side of the AI industry.
Models Are Not the Ones Competing; Applications Are Changing Gears
Jason, Unique Research
Using Unique Research's AI Top 100 data on global AI applications for November, Jason framed the AI market in 2025: the industry is moving from an installation phase of feverishly piling up infrastructure toward a deployment phase that solves real problems.
Jason argued that the real change lies not in how powerful models are, but in their capabilities beginning to reach practical use. He shared three core findings:
The model landscape is being reshaped: developers no longer focus exclusively on OpenAI. Anthropic and Gemini are becoming developers' preferred colleagues because of their stability in code collaboration, or vibe coding.
Video generation is industrializing: MovieFlow's 915% surge in visits reflects video generation's transition from technical spectacle to production lines with industrial-scale output capacity.
AI assistants are reaching the mass market: monthly active users of Ant Afu grew 664%, marking AI's genuine move from a community of enthusiasts into ordinary people's lives, where it begins solving frequent and essential needs such as interpreting medical-examination reports.
He emphasized that Chinese models are globalizing through a dual pattern of explosive domestic downloads plus overseas access points. Future competition will shift completely from cost-per-token terms—who is cheaper—to cost-per-successful-outcome terms—who can deliver results reliably.
Working with Fragmented Data to Become a Hidden Champion at Sea
Wu Bingkun, Founder and CEO of Zhongshu Xinke (众数信科)
Wu Bingkun explained how Zhongshu Xinke (众数信科) avoided direct confrontation with major companies over general models and instead chose shipping and marine engineering, a knowledge-intensive field with fragmented data. By using agent groups to extract knowledge, the team successfully built Wenyao, China's first foundation model for shipping and marine engineering. He emphasized that the real barrier in this industry is preserving veteran experts' tacit knowledge rather than simply piling up a general-purpose knowledge base.
For practical deployment, Zhongshu Xinke (众数信科) launched an AI Crew Member that uses integrated hardware and software to help shipowners manage equipment health and energy consumption, saving each ship genuine labor costs. The company is also advancing spatial intelligence and world models and will soon deploy unmanned vessels in Xiamen and elsewhere to collect floating marine waste and monitor the ocean environment. With reported revenue above 100 million and profit above 10 million (currency not specified in the source), Wu demonstrated that maximizing cost savings and efficiency in a vertical market is AI's best business model.
From SEO to GEO: Capturing the Trust Dividend in the AI Search Era
Cai Xiaoxu, Founder and CEO of Taimei AI (钛镁 AI)
Cai Xiaoxu identified a major shift in marketing logic during 2025: consumers are moving from traditional search toward asking AI, and the traffic gateway has changed. He introduced GEO, or generative engine optimization: brands must create content not only for consumers to see, but for AI to see, so AI understands and trusts the brand and can prioritize recommending its products when applications such as DeepSeek and Doubao answer users' questions.
He further broke down practical GEO methods. The core is satisfying E-E-A-T principles—Experience, Expertise, Authoritativeness, and Trustworthiness—by using high-quality professional content and authentic consumer-experience cases to strengthen the information-source foundation; authoritative data and endorsements from industry experts or institutions to establish a “digital business card” that AI trusts; and multimodal content arrangements to help AI understand better. Taimei AI (钛镁 AI) helped an infant-formula brand achieve an exponential leap, with average daily search volume growing nearly 120% month over month compared with the pre-optimization level. Cai emphasized that as content fragments and customer-acquisition costs remain high, using AI to intercept high-intent search demand is critical for brands seeking to overtake competitors around a bend.
The First Year of Agentic AI: From Selling Functions to Selling Results
Gao Zhou, Founder and CEO of Zhongju Intelligence, or Luna AI
Drawing on a quantitative-trading background, Gao Zhou brought the financial sector's logic of replacing traders with models into AI customer service. His Luna AI is not merely a passive response tool, but a super employee capable of proactively identifying customer intent, managing sales opportunities, and even perceiving emotions through WeCom (Enterprise WeChat) and other channels. He believes the core value of an AI Agent lies in solving data-structuring problems so that AI genuinely enters enterprise workflows.
Looking toward 2026, Gao offered a profound insight: the SaaS industry's business model will evolve from Selling Functions to Selling Results. Customers no longer care how advanced AI is; they care only whether it can produce ROI gains. He urged entrepreneurs not to chase the latest model parameters, but to build systems that customers cannot do without and cross industry cycles by delivering concrete results.
Using AI to Cut Real Costs in a Trillion-Scale Automotive Supply Chain
Zheng Zixi, CEO of SupplyOn
As a veteran professional managing the world's largest automotive-supply-chain platform, Zheng Zixi demonstrated the power of making an elephant dance. SupplyOn supports the lifelines of giants including Bosch and BMW and possesses deep stores of historical data. She noted that B2B applications—especially in manufacturing—have extremely low tolerance for AI errors and cannot spout nonsense confidently like consumer applications; they must rely on accurate historical transaction data and master-data management.
She demonstrated Aku, an AI employee that automatically initiates requests for quotations, compares supplier bids, and even provides negotiation advice. In a real case, AI directly helped an enterprise save 40% in the bidding process. Zheng emphasized that AI is not merely a cost-reduction tool but also a path to increasing revenue, provided that companies first complete the foundations of internal digitization. In supply chains that cannot stop operating, AI's value lies in freeing procurement specialists from tedious processes so they can make more strategic decisions.
An Engineer's Guide to Avoiding Traps: Go Global and Compete in More Valuable Markets
Li Biao, Founder and CEO of Zhidun Technology
As a serial entrepreneur with a technical background, Li Biao offered an exceptionally grounded account of painful lessons from his startup journey. He admitted learning costly lessons from blindly accepting orders and pursuing ineffective requirements, eventually discovering that pure technical outsourcing and highly customized projects are difficult to scale. After reflecting on the pain, he turned to overseas markets with greater willingness to pay and equipped his team with AI tools to unlock global business at extremely low labor cost.
Li demonstrated the team's internal AI public-opinion insight tool. With only several people, it replaces a traditional ten-person analysis team and identifies product pain points precisely by monitoring complaints across global social media in real time. His advice was highly practical: maximize the use of paid AI tools to improve efficiency; do not expand team size for vanity; remain small, focused, and agile; and find a high-margin survival niche in overseas markets.
The Truth Revealed by the Openday
Truth One: Do Not Run Alongside Foundation Models; Become the Hidden Master Sweeping the Floors of a Vertical Industry
Over the past two years, we placed too much faith in generality. In reality, major companies' general models resemble regular armies on an open plain: they are invincible, but there are places they cannot reach, including the gullies and ridges of vertical industries.
Wu Bingkun of Zhongshu Xinke (众数信科) chose an especially difficult road: shipbuilding.
What kind of industry is this? Knowledge lives in veteran experts' minds, data is scattered everywhere, and the environment is extremely harsh. Major companies are uninterested and unable to do the work. Yet that is precisely the opportunity. Wu Bingkun led his team in extracting this tacit knowledge and turning it into an AI Crew Member.
What was the result? One integrated intelligent machine placed on a ship can save the shipowner the labor cost of two people. It is that direct and simple, yet it enabled him to surpass 100 million in revenue (currency not specified in the source).
That is the democratization of knowledge.
What does this teach us? No matter how traditional the industry, if you can save customers real money, AI is the best leverage. Even dirty and exhausting jobs such as collecting floating marine waste and cleaning solar panels become good businesses when AI can perform them.
Truth Two: Traffic Logic Has Changed; Learn to Win Over AI
Marketers may all have felt anxious recently: Baidu can no longer find me, and users have gone to ask AI.
Yet the opposite side of anxiety is an enormous dividend.
When Cai Xiaoxu of Taimei AI (钛镁 AI) discussed GEO at the event, every operator in the room suddenly saw the opportunity.
Previously, we wrote articles to win over search-engine crawlers and users. Today, we must learn to win over AI. You must make DeepSeek and Doubao consider you authoritative and trustworthy.
Cai Xiaoxu did one thing for an infant-formula brand: he found experts to endorse it, found real users to write about their experiences, and fed the content to AI. As a result, the brand went from being found by only 1,500 people per day to 30,000 people. [This is a 20-fold level, or a 1,900% increase, whereas the earlier passage reports nearly 120% growth. The source does not explain whether these figures use different periods or definitions; both original claims are retained without reconciling them.]
This is more than a technique; it is a dividend from understanding. While others still lament that traffic is becoming more expensive, if you know how to make AI your recommendation agent, you have already won at the starting line.
Truth Three: Stop Selling Functions and Start Selling Results
Gao Zhou of Zhongju Intelligence said the SaaS industry has changed completely. Previously, software was sold by Selling Functions: I have this function, so buy it. In the future, we must use Selling Results.
Customers do not care whether you use Llama or the GPT-4 model. They care only about this: can you help me recover 25% of lost orders? Can you help me increase conversion by 72%? These are illustrative customer questions in the original article, not measured results from a named case.
That is the cruelest and most fascinating aspect of 2025: do not look at the advertisement; look at the outcome.
Zheng Zixi, referred to in the source by the nickname 茜茜, who manages the world's largest automotive-supply-chain platform, is best qualified to speak on this point. Manufacturing tolerates no AI hallucination. Her AI employee Aku does not write poems or make pictures; it does one thing—helps companies negotiate lower bids. One order can directly save 40% of the cost.
Customers cannot do without you only when you help them earn the money back. That is more compelling than any financing story.
Closing Thoughts
AI has finally changed from mysterious high technology into electricity, water, and gas, illuminating households everywhere.
In 2025, there may be fewer earth-shaking disruptions, but countless applications such as AI crew members, AI procurement specialists, and AI customer-service agents will quietly change how the world operates in overlooked corners.
This is also Li Biao's advice to everyone as an entrepreneur with an engineering background: do not chase scale and do not be vain. Equip yourself with AI. Even with only two or three people, if you can solve one genuine problem, you are a super individual of this era.
This is the genuine demand of 2025: fewer concepts and more results; less anxiety and more action.
If you also want to find your position in this wave, ask yourself: for whom can I use AI to solve even one specific, small problem?
Registration Is Now Open for the Next Unique OpenDay
📅 Time: Friday, January 23, 2026, 13:30-18:00
📍 Location: Main venue, floor 1 of building 7, Jingronghui, Chengdu High-Tech Industrial Development Zone
Organizers: Unique Research and the Sichuan Artificial Intelligence Research Institute
Co-organizers: Volcano Engine and Akamai