This Tuesday, three AI large models, Genie, GPT-OSS, and Opus-4.1, were released, and today GPT-5 was also launched. The wave of AI innovation continues to surge. Although tech media reviews generally believe that GPT-5's performance does not reach the level of AGI, the main improvements are in supplementing the shortcomings of Vibe Coding and increasing the accuracy of solving complex mathematical and physical problems. However, the optimistic sentiment in the financial markets and the AI industry regarding the "AGI moment" has not been significantly affected. Judging by the current advancements in Scaling Law, Agentic learning, and reinforcement learning, along with the views of industry leaders like Musk and Dario Amodei, superintelligent agents are expected to emerge between 2026 and 2027. The financial market will speculate before the arrival of the AGI moment, so we need to proactively search for related targets and get involved on the left side to seize 🫴 this new generational investment opportunity. Given that the integration journey between the US stock market and the crypto market is beginning, rather than waiting for AGI-related targets to go through various cumbersome Tradfi IPO procedures to open up to secondary market players, it is better to directly look for highly relevant participatory targets in the crypto market. Under the premise that the existing training paradigms of AI large models remain unchanged, we can find one optimal target from four dimensions: data, large models, computing power, and AGI Agents. –-Data Data, or corpus, is the foundational material for training AI large models. The reason GPT-3.5 achieved a breakthrough in performance as a large predictive model is due to the introduction of a large amount of Reddit and X Twitter corpus. Nowadays, issues such as the impending exhaustion of human-generated databases, AIGC pollution and toxic feedback loops, and the preprocessing and management of enterprise/personal private data are constraining the emergence of AGI. Therefore, from the data dimension, compared to Chainbase and Ocean Protocol, which specialize in providing on-chain data to supplement training corpus, Sahara AI, which utilizes token economics to incentivize the production of high-quality, multidimensional, structured corpus collaboration networks, is undoubtedly the best target. –-Large Models Currently, large models are not only competing in performance but also starting to compete in use cases and usage costs. For example, GPT-OSS, released by OpenAI before GPT-5, is a cheap and user-friendly open-source large model. In the AGI model segmentation track, the Artificial Superintelligence Alliance ($FET) and Bittensor ($TAO) have deep technical foundations and brand histories, but the former's project operation has a sense of "retirement after success," making Bittensor ($TAO) a better choice. Additionally, although some emerging projects have not yet reached the TGE stage, they are worth our close attention. Alternative high-valuation primary market projects include SentientAGI, which is dedicated to building AGI commercial ecosystems, Nous Research, which specializes in decentralized AI training, and Prime Intellect. –-Computing Power Computing power is the oil of the AI era; whoever controls the distribution of computing power controls the pricing power, which has become a market consensus. The underlying logic of decentralized GPU networks (GPU DePIN) is AI explosion → extreme scarcity of computing power → centralized cloud computing monopoly pricing → DePIN is an inevitable trend. Currently, the main optional targets for GPU DePIN include Aethir, which possesses its own large-scale GPU computing power, has enterprise-level computing support capabilities, and is building a prosperous and diverse application layer ecosystem that includes AI Agents, Web3 games, and RWA protocols. Moreover, @AethirCloud has quietly achieved an ARR of $156 million. What does this number mean? Over 99% of Crypto projects do not earn this amount throughout their entire lifecycle 🤯 –-AGI Agent The AI Agent frameworks Virtual and AI16Z, which have become new coin launchpads, and the AI Agent of the garbage X post-publishing robot have been eliminated by the times. New AI Agent paradigms more suitable for AGI intelligence levels are emerging, such as the Agent Social networks being explored by @ChainOpera_AI and @Infinit_Labs. The new AI Agent paradigm pursues true usability, achieving real cost reduction and efficiency improvement while generating economic value in product development, financial trading, and self-media content writing. Currently, neither of these two projects has a TGE; whoever has a TGE first will be prioritized.