Original · Unique Research · 2026-01-25
Historical edition: This complete textual event report was published on January 25, 2026. “Today,” current roles, institutional partnerships, product capabilities and forecasts retain the original report's historical perspective, not independently verified current status. The 80% labor-efficiency gain, 800-person production scale and 300 participants appear in the original section headings; the prose does not provide a measurement methodology for the 80% claim. Chinese organization names without an established English form are rendered descriptively or transliterated. Nominations and other launch announcements are historical, not current calls to apply. Images and image-only material are outside this text-only edition.
When Pan-Entertainment Meets AIGC: A Chengdu OpenDay Focused Only on “Real Closed Loops”
It is difficult not to relax on a Chengdu afternoon, but today's relaxation carried an added sense of urgency: AIGC is blending the boundaries among the content industry, the tools industry, and global business. OpenDay, an offline AI salon jointly initiated by Unique Research and the Sichuan Artificial Intelligence Research Institute, concluded successfully in Chengdu High-Tech Industrial Development Zone. The event gathered AI companies, AI builders, AI creative professionals, industry investors, and ecosystem partners. People with different identities sat in one space and seemed to have come for the same reason: to set aside valuation narratives and explain clearly how to find paying users, deliver results, and build sustainable businesses.
Unique Research has always defined Unique Research OpenDay as a “stage for people on the same wavelength”: it pursues truth rather than spectacle. Today's theme was equally direct—when pan-entertainment meets AIGC and AI applications meet overseas markets. More than ten representatives of Chengdu's AIGC and global-business community used open-mic talks and roundtables to break down the path from product to business across four forms—tools, content, services, and platforms: which sector to choose, how to build a product, where to find customers, how to establish a brand, and most importantly, how to turn one delivery into long-term compounding returns.
Opening with the “Value Deployment Phase”: Moving the Conversation from Technical Spectacle Back to Returns
The event's opening note was not “the model has become stronger again,” but “the industry's phase has changed.”
In the opening remarks, a representative of the Sichuan Artificial Intelligence Research Institute defined the present as a transition from an “installation phase” that chases parameters toward a “value deployment phase” focused on cost reduction, efficiency gains, and commercial returns: technology no longer ends with amazement, but must land in measurable benefits. Chengdu's advantages were also stated clearly—university talent, living and experimentation costs, and deep cultural and creative roots—which give the city the potential to become both a low-cost base and a creative hub for AIGC companies going global. Yet whether it can travel far still depends on connecting ecosystem collaboration and the talent pipeline.
Zhao Liang, a partner at Unique Capital, then quickly introduced the Keynote speakers and salon guests, establishing the day's rhythm: every speaker should try to answer the same question—how does your product ultimately collect revenue?
Grounding the Discussion in Data: Using the “AI Top 100” to Return to the Real World
Huan Jiachen, head of research at Unique Research, then used AI Web data from December 2025 to argue that the main thread in AI applications is shifting from “who controls the gateway” to “who can deliver concrete tasks and embed itself in production processes.” Using the ranking's structure, he emphasized that concentrated growth has spread from conversational products alone into utility categories, especially coding assistants, video and image creation tools, and agents, which continue producing more stable user curves. Products that maintain more than 20% month-over-month growth even from a base of millions of monthly active users tend to occupy content production, emotional companionship, efficiency gateways, agent scenarios, and related areas. More specialized and monetizable needs are absorbing incremental Web demand. The December ranking sends a clear signal: competition in 2026 will compare not only model capability, but ecosystem integration, scenario deployment, and mindshare among developers.
From 2023 to 2026: The Second Half of AIGC Competes on “More Complete Content Delivery”
The first presentation to truly ignite the room came from Ma Fei, CEO of SeaArt AI.
He described product evolution from 2023-2026 as something closer to a commercial climate: 2023 was an introductory period for text and images, with leading positions in companionship and pure-image products largely established; in 2024, images expanded into entertainment, and filters and light-entertainment tools began reaching scale; he called 2025 the year multimodality truly accelerated, with the leap in video making 5-10 seconds a new norm; and his assessment of 2026 was sharper—coding agent systems may let creation skip part of the traditional toolchain, using natural language to control pixels, rhythm, music, and interaction directly. Content will experience another explosion, but competition in the second half will no longer center on whether something can be generated. It will center on who can deliver differentiated, more complete, and more consumable content.
He also shared specific experiments from the team's early exploration: combining a multimodal agent with marketing and Web coding to create more end-to-end forms of delivery, moving from “generating assets” to “generating usable interactive content.” These experiments are exciting not because they sound advanced, but because they address what everyone cares about today: whether delivery is complete enough to support charging customers.
Doubao's Pan-Entertainment Path: Social Interaction, Games, Short Dramas, and Animated Comics Are Being Rewritten by “Generative Production”
The next speaker was Chen Jie, senior director of foundation-model products at Volcano Engine. He divided pan-entertainment into three curves undergoing structural change:
social companionship is advancing from interaction to emotional connection, with multimodality making characters into entities that can be designed and managed continuously; short dramas and animated comics are shifting from filming toward generative production as the text-to-storyboard-to-image-to-video-to-synchronized-audio-and-visual chain begins operating end to end; games are advancing from scripted experiences toward “infinite interaction,” where AI-NPC characters, real-time generation, and dynamic levels prevent content from remaining a one-time consumable.
He presented Doubao's practical thinking around role-playing, multiplayer group chats, voice experiences, long-term memory, and other capabilities. He also demonstrated ideas involving interactive stories, conversion between illustrations and animated images, storybook output, character-image generation, costume changes, storyboards, and plots. On the gaming side, experiments spanning story dialogue in Justice Online, multi-person conversational characters, and generation of the next level and scene all point to the same trend: AIGC is turning content from “output” into “continuous generation,” and from “finished after consumption” into something that grows longer with use.
“The Starting Point of Infinity”: A Foundation-Model Company Discusses Post-Training, Agent Systems, and Organizational Intelligence
When the topic shifted from applications to models, the atmosphere in the room became noticeably quieter. Fu Qiang, Technical Staff at Moonshot AI, spoke under the title “the beginning of Infinity” about post-training methods, foundation-model companies' long-term perspective, and possible future forms of the Agent.
By comparing supervised fine-tuning (SFT) and reinforcement learning (RL), he made the distinction between “following instructions” and “the ceiling of intelligence” intuitive: SFT resembles teaching a model to solve problems through fixed steps and suits imitation and formatted output; RL resembles letting a model discover rules through feedback, which costs more but has a higher ceiling. He then turned to the Agent. Chat resembles a “brain in a vat,” while an Agent is a “brain + hands and feet + notebook + supervisory feedback” that delivers results on long-horizon tasks. The Agent of 2026 may become organized and form the beginnings of a kind of “organizational intelligence.”
The presentation did not deliberately manufacture emotion, yet it gave the room a rare sense of certainty: however lively applications become, underlying training paradigms, feedback mechanisms, and long-horizon task capability will still determine how far the next stage can go.
“China's Sora” and Practical Video Generation: Lowering the Barrier to Creation Again
The next speaker was Wang Chuan, vice president of ShengShu Technology. He began with the evolution of multimodal-model companies and focused on practical video generation: across duration, performance, cinematic language, semantic understanding, and editing, video models are turning “usable” into a standard component.
He shared multiple directions closely connected to pan-entertainment production: interactive templates resembling short-video effects, MV generation from combinations of images and music, editing that advances from image references to video references, subject modification and local editing, and even capabilities involving body-motion references and synchronized audio-video output. These may sound like “features,” but for the content industry they mean something more brutal and real: capacity will continue rising, costs will continue falling, and victory will depend increasingly on creativity, narrative, and commercialization paths.
An “Alliance” Brings Universities, Enterprises, Capital, and Scenarios into One Network
An important event milestone was the unveiling ceremony for the AIGC Industry-University-Research Alliance. It is not intended as a social-media group photo, but as a means to connect the last mile from laboratories to industrial applications, coupling university knowledge, enterprise scenarios, capital, and market demand more efficiently and cultivating multidisciplinary talent that understands both technical language and commercial logic.
Representatives initiating the alliance included She Yanjun, vice president of the Sichuan Artificial Intelligence Research Institute; Wu Wei, founder of Unique Research; Wu Xuefeng, director of the Sichuan Digital Intelligence Culture and Tourism Professional Committee; Gao Zhongli, vice dean of the Chengdu Academy of Fine Arts at Sichuan Conservatory of Music; Xiang Dong, dean of Digital Media and Creative Design at Sichuan University of Media and Communications; Ma Fei, CEO of Chengdu SeaArt Interactive Entertainment Technology; Sun Tingyan, founder of Chengdu Xinshijue Culture Communication; Li Yonghong, founder of Dacheng Chanjing (大成产经); and Yang Xiaofeng, founder of Huanzhou AI.
When the countdown ended and everyone pressed the launch button together, the applause was remarkably synchronized. Everyone understood that the content industry's true shortage has never been inspiration, but a sustainable supply of people who understand both technology and narrative, and both narrative and business.
Nominations Open for the Chengdu AI 30: Turning “Shared Frequency” into a Sustainable City Narrative
Nominations for the Chengdu AI 30 ranking then formally opened. Its timing was no coincidence: when a city begins systematically identifying, recording, and supporting its emerging forces, entrepreneurs need not rely only on accidental hits or isolated resource connections. The ranking is not the endpoint but a capacity for city narrative, helping more people see the verifiable AI businesses emerging in Chengdu.
“Cost” Is Both Barrier and Weapon: Three Paths for Inference, Deployment, and Scale
As the discussion moved from “can the product be built?” to “can scaling work?”, Dr. Zhang Lu of Akamai focused on a question no one can avoid: how to use models more economically. He compared the costs and applicable scenarios of three approaches—calling an API, using a cloud provider's GPU servers, and building private servers—and offered an engineering-oriented judgment: once workloads become stable and scale increases, the stable portion can move to private GPU infrastructure while API services handle fluctuations, potentially reducing overall cost substantially. Different deployment thresholds also exist for speech, image, and large-language-model scenarios. He introduced Akamai's advantages in inference and edge nodes: deploying inference workloads closer to users reduces latency and increases retention. This easily overlooked infrastructure calculation determines whether an app can ultimately make money.
An AI-Native Company's Organizational Experiment: 80% Higher Labor Efficiency and Jobs Recast as “AI Coordinators”
Li Yuanyuan, co-founder of Mobvoi, had many people nodding repeatedly during the latter part of her presentation. She did not discuss “using AI tools to improve efficiency,” but how a company with established operations and processes integrates AI into its organizational structure itself. Across experimentation, symbiosis, and efficient collaboration, tools are required to enter every meeting and collaboration, workflows and context are accumulated systematically, and role names change accordingly: engineers become “full-stack engineers,” communication roles become “human communicators,” and colleagues increasingly resemble “AI coordinators” who each play many roles.
The most moving part of the presentation was its honesty: AI executes, while humans provide meaning. The difficulty in organizational change lies not in the tools, but in persuading people to let a new collaborative paradigm rewrite how they work.
The Road to AGI May First Pass Through business-to-business “Disorder and Hallucination”
Chen Hongyu, co-founder of Xingzhe AI, then discussed possible paths toward AGI at a more macro level: AI's development resembles the evolution of human society, moving first through communication, then tools, then efficiency, and ultimately perhaps toward organized collaborative systems. He also confronted business-to-business challenges directly—disordered workflows, hallucinations, and the lack of industry common sense in both general and vertical models—and proposed combining a “cognitive operating system” with industry data, compliance, and structured accumulation to build more controllable output for industries.
When he compared his role to a “smart grid”—not building a general foundation model, but delivering AI capabilities to industries stably, safely, and efficiently—many enterprise-service professionals in the room wore expressions suggesting, “Finally, someone has reached this point.”
A Production Machine of 800 People: Animated Comics and AR Short Dramas Turn Content into a Supply Chain
The penultimate presentation came from Wang Shubo, founder and CEO of Judian Short Drama and Lingju Animation. He described in highly industrial terms the company's path from live-action short-drama hits to overseas distribution and then into animated comics with AI: team size, shifts in production models from 1+6 to 1+2 and 1+1, how AR tools move part of filming onto computers, and how capacity doubled over half a year. These details reveal that AI content is not a game of inspiration, but a production line that can be optimized, managed, and scaled.
He also reminded the industry that content must ultimately return to the essence of storytelling. As supply becomes more abundant, business models and the rules of survival—not temporary excitement—will determine life or death.
Chengdu's Global Breakthrough Will Ultimately Be Decided by “Taste, Creativity, and Ways of Collaborating”
The final roundtable, titled “How AIGC Empowers the Content Industry and Chengdu's Path to Global Breakthrough,” was moderated by Song Kaifa, co-founder of Chengdu-based VMEG. Guests included Xu Long, head of AIGC at Wukong AI; Yang Xiaofeng, founder of Huanzhou AI; Sun Tingyan, founder of Chengdu Xinshijue AIGC; Zhu Wenxu, general manager of Chengdu Aimenweisi (成都艾门韦思); Luo Linfan, general manager of Yuguang Chuangying; and Chen Ye, head of a teaching and research section at the Chengdu Academy of Fine Arts, Sichuan Conservatory of Music.
The roundtable felt distinctly Chengdu: no rush to conclusions, yet every sentence focused on execution. Regarding AI's impact on live-action dramas, the discussion leaned toward integration rather than replacement. Regarding education and talent, both enterprises and universities identified the same pain point—people who genuinely understand the industry, aesthetics, and workflows remain scarce. Regarding excessive competition, participants translated “differentiation” into concrete actions: identify content's consumer value, turn creativity into sustainable production, use partners to fill capability gaps, and find expressions between local IP and global narratives that markets can validate. When Chen Ye said universities hoped to teach students to combine aesthetics with AI and prevent creation from becoming “AI slop,” the room laughed, but the laughter was complex: both humorous and cautionary. Tools will keep growing stronger, but whether a work earns payment may ultimately still depend on human judgment about beauty and stories.
A “Speed-Matching Conversation” Among 300 Like-Minded Participants: Bringing the Methods on Stage into a Handshake Before Departure
If the presentations provided “paths,” the connections offstage provided “speed.” Participants checked in through a mini program and used “speed matching” and an information wall: seeking technology, creativity, investment, and collaboration. Writing needs this directly made it easier to find the right people quickly. For many participants, the day's most valuable moment may not have been a memorable quotation, but the confirmation after exchanging business cards before leaving: the missing piece of my puzzle was sitting next door.
Epilogue: Chengdu's Answer Is Not “Louder,” but “Better at Delivering”
As the event ended with applause and group photographs, it left behind not spectacle but clarity: AIGC is reshaping pan-entertainment's global value chain, yet the closed-loop capability to deliver, reuse, and charge will always be what survives cycles.
Unique Research and the Sichuan Artificial Intelligence Research Institute have established a close strategic partnership. As a provincial artificial-intelligence research institution approved by the Sichuan Provincial Department of Science and Technology, the institute was jointly established with support from the Chengdu High-Tech Industrial Development Zone Management Committee, University of Electronic Science and Technology of China, Sichuan University, and other parties, providing Unique Research with academic resources and support from the local industrial ecosystem. The two sides continue advancing deep industry-university-research integration, jointly organized this OpenDay event, co-founded the AIGC Industry-University-Research Alliance, and will jointly publish authoritative industry rankings including the Chengdu AI 30. Together, they aim to move Sichuan from an “industrial highland” toward an “AIGC innovation source” and build a more credible and externally influential exchange and collaboration platform for local AI practitioners.