Qué es $DOGE

Dylan, the founder of SemiAnalysis, revealed that Anthropic may be holding a more powerful model than the publicly available version, but it has not been released due to regulatory and safety pressures. He mentioned that Anthropic's Mythos has already been trained but has been delayed for months, and Mythos 2 is reportedly also completed but remains unreleased. The publicly available Fable has incorporated numerous safety classifiers with very strict access restrictions. Dylan believes that what is truly concerning is that although these stronger models are not available to the public, Anthropic is likely still using them internally to continue training the next generation of models. This means there may already be a significant gap between the AI capabilities seen by the public and the actual capabilities held within the lab.

On August 18, according to monitoring by Ai Yi, a suspected LAB insider address 0x0d9…751d0 transferred 9.1 million LAB tokens, worth approximately $720,000, to 10 new addresses three hours ago. Currently, the receiving addresses have not shown any further transfer or selling activity. It is reported that this address was previously identified as a LAB whale address, and market participants are concerned whether this recent distribution of transfers indicates a possibility of continued sales in the future. The current market capitalization of LAB is approximately $36.85 million.

On August 18, according to Phoenix Finance, Professor Wang Jiange, chief researcher at Japan's NTT Data Group, stated that there is a significant industrial bubble surrounding Nvidia, and the current monopoly with high profits is unlikely to be sustained in the long term. Regarding the lifespan of the Nvidia industry bubble, Wang Jiange made a clear prediction: 'The bubble will burst within three years, and the current biggest bottleneck is power supply.' He explained the root cause of the bubble formation: the existing black-box large models lack efficient mathematical description tools, and the industry has fallen into the misconception that 'the larger the model, the stronger the competitiveness,' completely detached from physical constraints. 'To describe an apple falling, Newton's mechanics only requires three parameters, while existing large models need billions of image data. Once a new AI mathematical theory is established, the demand for computing power could decrease by millions of times, and Nvidia's current business logic will no longer exist.' In the context of the bubble bursting, the storage sector is expected to encounter long-term opportunities. Whether the future mainstream is electronic AI chips or photonic AI chips, data storage is a necessity. Companies like Lanqi and Changxin will not be significantly affected by the drastic fluctuations in computing power cycles, and their industry growth cycles will far exceed those of computing power manufacturers like Nvidia.