最新报道：According to Bijie.com, artificial intelligence (AI) engines have long been plagued by hallucinations, causing them to fabricate answers out of thin air. In fact, studies have shown that newer engines have error rates as high as 80%. Can this problem be solved by integrating into the cryptocurrency space? According to a study completed by AI Foundations on June 29th, researchers found that 16 engines had hallucination rates exceeding 15%, with Gemini 2.0 Flash and GPT 4.1 being the highest. A separate test report released by OpenAI showed that its o4-mini model had an error rate as high as 79%, a significant figure considering the number of queries these models answer. To address this issue, layer 1 blockchain startup Nuklai has launched a new AI engine called Nexus, claiming it can "bring together different types of information in a unified format." Backed by AI and cryptocurrency giants such as IO.net, Filecoin, and Fetch, the project deploys resources both on-chain and off-chain to mitigate bias, errors, and other inconsistencies in AI models. According to Coingecko, AI agents account for a significant portion of the $28.25 billion market capitalization. While a market cap of nearly $30 billion may seem modest compared to the broader $3.7 trillion cryptocurrency market cap, the sector has demonstrated significant growth potential. Compared to the same period in mid-2024, the market capitalization of the AI-to-cryptocurrency sector has increased sevenfold, from just $4.04 billion to over $28 billion within a year. AI agents, in particular, saw their market capitalization peak at nearly $17.50 billion in January 2025. However, it has since slumped, barely exceeding $6.00 billion as of July 2025. This suggests that AI agents are in need of a boost in the cryptocurrency sector. As early as December 2024 and January 2025, the AI-to-cryptocurrency sector saw a surge in collaborations, with cryptocurrency giants like Ripple teaming up with AI platforms like Atua to build AI solutions for the XRP ecosystem. Now, the Nexus project is attempting to revive this trend by bridging AI and cryptocurrency, creating a new model that can further integrate the two fields. Through Nexus, Nuklai aims to build a network of over 80 partners across the AI and cryptocurrency sectors, including IO.net, Fetch, and Filecoin. This partner network aims to provide Nexus with access to powerful models without having to manage infrastructure or pay for expensive enterprise application programming interfaces. IO.net Chief Business Development Officer Tausif Ahmed stated that the project's collaboration with Nexus aims to leverage decentralized inference to improve the practical application of AI. "By integrating large-scale language models hosted by io.net directly into the Nexus setup process, developers can instantly access powerful models without having to manage infrastructure or pay for expensive enterprise APIs," Ahmed said. By deploying the Model Context Protocol—an engine that compiles data sources such as databases, files, APIs, and cloud services into its servers—the developers claim it will be able to provide users with answers accompanied by explanations of how each result was generated. Nuklai founder Matthijs de Vries stated that most tools that use artificial intelligence as their models often sound smart on the surface but break down when needed most. "With Nexus, we're changing that. It connects directly to the data, shows you the source of the answer, and finally solves the problem of illusion that everyone is dealing with," de Vries said in a press release shared with crypto.news. Nuklai claims that the problem with today's AI models is that they fail to display provenance and tend to fabricate answers. It describes itself as an "AI-agnostic" system, meaning it supports all major LLMs without being locked into a specific vendor. The project's technical foundation relies on NXSQL, a computational language that works with virtual systems, where each piece of information is organized into its own pool. This allows users to obtain answers from disparate sources in a manner that it claims is systematically organized. However, research indicates that the main issues extend beyond disorganization or limited access to training data. In an AI Foundation study, other reasons for AI presenting incorrect information were related to misunderstanding user prompts and a tendency to focus more on producing fluent, clear, and concise writing than presenting accurate facts. Another key factor highlighted by the study involves the training deadlines of most LLMs, which make them slow to update on current events or emerging trends. While much work remains to be done in AI improvements, Nexus' modular design, open integration, and integration between blockchain and AI reflect a thriving field with potential for growth. The emergence of greater collaboration and shared resources could pave the way for greater integration between cloud computing, blockchain, and AI models in the future.