At the Industrial Ecology Forum of the 2026 World Artificial Intelligence Conference (WAIC) on July 18th, ABLE DIGITAL (HKEX: 02687), as a benchmark enterprise in the knowledge technology sector, was invited to speak and officially launched its core achievements in knowledge technology: the "Source Graph" knowledge base and a new Harness engine. This was not a routine technical release but a systematic response to the core pain points of the AI + knowledge industry: the bottleneck lies not in models, but in knowledge. Notably, during the "University Presidents Dialogue" session at the same conference, seven academicians and presidents from top universities including Tsinghua, Peking University, Shanghai Jiao Tong University, Huazhong University of Science and Technology, Harbin Institute of Technology, and HKUST (Guangzhou) discussed the theme "When Machines Can 'Teach', What is the Role of Universities?" The recurring keyword was "knowledge." The launch by ABLE DIGITAL directly addresses this very question.

The Academic Challenge and a Direct Response

During the forum, Peking University President Ding Kailing noted that low student engagement is not due to lack of effort, but because the traditional "knowledge-dumping" teaching method is ineffective when AI can instantly retrieve any information. Huazhong University of Science and Technology President You Zheng was more direct: AI can serve as "scaffolding," but not as a "crutch." These discussions point to a common direction: in the AI era, the core mission of universities is shifting from "how to teach knowledge" to "what knowledge to teach"—and who provides this knowledge. Delving deeper, a more practical question emerges: if machines are to "teach," how can they teach accurately? ABLE DIGITAL offers a clear judgment—the real pain point for the industry currently is not the model, but the knowledge. The training data for general large language models mainly comes from public internet data, but the knowledge that truly constitutes disciplinary barriers—textbook systems, experimental data, decades of accumulated teaching experience, and professors' expert judgment—is precisely what cannot be retrieved from the internet. For scenarios with extremely high precision requirements, such as physics experiment procedures, engineering design standards, or medical diagnostic logic, factual hallucinations are not "flaws" but "fatal errors."

Source Graph: Ensuring Traceable AI Reasoning

ABLE DIGITAL systematically disclosed its AI's underlying architecture—the "Source Graph" knowledge base—for the first time at WAIC. The "Source" in the name points to a core requirement: traceability. On this foundation, the AI's reasoning process is no longer a "black box"; every step of the deduction can be traced back to specific academic literature and knowledge nodes. The company serves nearly 2,000 higher education and academic research institutions, covering all academic disciplines. All knowledge undergoes source screening and supports full-chain traceability. This data originates from teaching environments, experimental settings, and front-line research, featuring high update frequency, deep structuralization, and dynamic iteration. Essentially, the company is systematically transforming the tacit disciplinary knowledge originally scattered across textbook systems, experimental standards, lesson plan designs, and teacher experience into a structured knowledge infrastructure that machines can read and utilize.

Harness Engine "1+N+1+N": Productizing Knowledge Infrastructure

Beyond its analysis, ABLE DIGITAL also launched a new Harness engine—"1+N+1+N." The first "1" is the Source Graph knowledge base, which deconstructs disciplines into four overlapping layers: knowledge point graphs, competency graphs, experimental graphs, and assessment graphs, forming a discipline-level "knowledge operating system." The "N" represents multi-scenario paradigm modules covering different business scenarios such as research, teaching, industry, and experimentation. The second "1" is the cultivation platform, managing the entire talent development process. The final "N" stands for discipline-specific customization for deep adaptation to particular fields. The delivery method is crucial: it is decomposable, combinable, and accessible on-demand. Users do not need to overhaul existing systems but can adopt a modular integration approach, connecting gradually by discipline and scenario, integrating as they mature. This means "1+N+1+N" is essentially a commercializable knowledge infrastructure delivery solution—with discipline-specific professional knowledge bases and vertical models as its foundation, featuring characteristics of scalable replication and decreasing marginal costs.

At WAIC 2026, ABLE DIGITAL presented not just a product launch, but a systematic solution targeting the pain points of AI for Science. Discipline-level knowledge data forms the foundation, proprietary AI capabilities serve as the engine, and a nationwide service network provides the delivery channel. The combination of these three elements moves "knowledge infrastructure" from concept to engineering implementation. As the Source Graph and Harness engine are deployed in more universities, the commercialization trajectory of the company's knowledge technology business is expected to accelerate, building a thicker and more substantial industrial moat for long-term value.