ChainCatcher news, according to an analysis by Qihoo 360 Xlab, the large-scale cyber attacks that caused three outages on the X platform are attributed to the same main botnet that attacked DeepSeek during the Spring Festival. This botnet is known as the Mirai variant botnet RapperBot, which belongs to a "professional thug" type of attack organization.Qihoo 360 Xlab discovered that the attack times perfectly matched the downtime of the X platform, primarily concentrated between 10 PM on March 10 and 2 AM on March 11 Beijing time. This botnet has been active for years, attacking hundreds of targets daily on average, with thousands of commands issued during peak periods. The attack targets are distributed across regions including Brazil, Belarus, Russia, China, Sweden, and others. The scale and intensity of this attack directly led to three significant service interruptions on the X platform, with a peak of 20,538 users reporting issues on March 10, Eastern Time.

ChainCatcher news, according to IBTimes, the founder of game studio Dizzaract, Ilman Shazhaev, stated in an interview that DeepSeek's low-cost AI development approach is "encouraging" and could become an important catalyst for blockchain development. He pointed out, "If high-performance AI models can be obtained at low cost, more decentralized systems will be able to utilize them at various levels."Shazhaev believes that DeepSeek's open weight design and two-stage reinforcement learning training method signify a true shift in AI development philosophy. He predicts that this model will have a significant impact on blockchain gaming, particularly in promoting smarter in-game economies.Although OpenAI CEO Sam Altman has accused DeepSeek of secretly using data from American AI giants to train its AI assistant, Shazhaev believes that a shift is occurring, especially among developers in the field of cryptocurrency and AI integration. He stated that budget constraints have always been a barrier in blockchain development, making models like the DeepSeek chatbot an attractive option.The AI model launched by DeepSeek in January of this year caused a stir in the tech industry, leading to a decline in related tech stocks and AI token prices. U.S. President Trump subsequently issued a statement calling for American tech companies to "wake up" and learn from this AI technology that does not require billions of dollars in development costs.

ChainCatcher news, according to the Bitcoin office of El Salvador, President Bukele recently met with a16z co-founder Marc Andreessen and benahorowitz.eth. The topics discussed included: investment opportunities for El Salvador as an emerging regional tech hub focused on free technology, the AI landscape in the post-DeepSeek environment, and open-source versus proprietary models in AI, among others.

ChainCatcher news, according to the Cybersecurity Bureau of the Ministry of Public Security, counterfeit DeepSeek websites and Trojan viruses have appeared online, causing great distress and security risks to the public.It is reported that a criminal gang launched a blockchain project called "DeepSeek," claiming to be a distributed ledger project developed based on DeepSeek technology, attracting many investors to participate in crowdfunding. After raising a large amount of funds, the team disappeared. Recently, many people have been pulled into the so-called "DeepSeek official group," where some claim to provide "beta testing qualifications" or "advanced courses," and created a virtual currency called "DeepSeekCoin," attracting a large number of investors to purchase, resulting in all the funds invested by investors being lost.The cyber police remind everyone to take precautions, download the genuine DeepSeek from the official website and mainstream application markets, and be vigilant against various security risks to avoid the possibility of downloading counterfeit programs from the source.

ChainCatcher news, according to DeepSeek's announcement, on the fifth day of Open Source Week, its Fire-Flyer file system (3FS) has officially been open-sourced. As a high-performance parallel file system, 3FS can fully utilize modern SSDs and RDMA networks to achieve high-speed data access, enhancing the efficiency of AI model training and inference.Key performance indicators of 3FS:Achieved a total read throughput of 6.6 TiB/s in a 180-node clusterReached a throughput of 3.66 TiB/min in a 25-node GraySort benchmark testSingle-node KVCache query peak throughput exceeds 40+ GiB/s3FS adopts a separated architecture, supporting data preprocessing, dataset loading, checkpoint storage/recovery, embedded vector search, and inference KVCache queries, with strong consistency semantics. DeepSeek also launched the Smallpond data processing framework to further optimize 3FS's data management capabilities.

ChainCatcher news, according to Jinshi reports, Deepseek announced the open-source optimization of parallel strategies on the fourth day of Open Source Week, including DualPipe------a bidirectional pipeline parallel algorithm for computing-communication overlap in V3/R1 training; EPLB------an expert parallel load balancer for V3/R1; and analysis of computing-communication overlap in V3/R1.

ChainCatcher news, DeepSeek announced today that starting from now, it will launch a staggered discount activity during the night period from 00:30 to 08:30 Beijing time every day. During the event, the API call prices will be significantly reduced, with DeepSeek-V3 dropping to 50% of the original price and DeepSeek-R1 as low as 25%. The official statement indicates that this move aims to encourage users to make full use of the idle night hours and enjoy a more economical and smooth service experience.

ChainCatcher news, according to Jinshi reports, Nvidia will announce its fourth-quarter earnings after the U.S. stock market closes on Wednesday. As the world's second-largest company, Nvidia needs to prove to investors that the demand for its next-generation Blackwell chips remains strong and is not impacted by the DeepSeek AI model. Analysts expect Nvidia's revenue for this quarter to grow by 73%, reaching $38.2 billion, compared to approximately $20 billion in the same period last year.Here is Wall Street's interpretation of Nvidia's upcoming earnings report:Mizuho Securities analysts expect Nvidia to release an "in-line" earnings report, but there may be "growing pains," as data center revenue for the April quarter may stabilize.Wedbush Securities believes Nvidia will perform strongly again and "clearly exceed expectations and raise guidance," with their survey indicating that individual AI company deployments have not slowed down due to DeepSeek.Bank of America expects Nvidia's data center revenue in 2025 to grow by 60% year-over-year and points out that the GTC conference in March is the next important catalyst.Bloomberg Intelligence states that Nvidia is likely to meet fourth-quarter revenue expectations and raise first-quarter expectations against the backdrop of significant growth in Blackwell GPUs.

ChainCatcher news, Deepseek announced the open-source of DeepGEMM on the third day of Open Source Week. DeepGEMM is a library designed for concise and efficient FP8 general matrix multiplication (GEMM), featuring fine-grained scaling capabilities as proposed in DeepSeek-V3. It supports GEMM for both standard and mixture of experts (MoE) groups.The library is written in CUDA and does not require compilation during installation, compiling all kernels at runtime using a lightweight just-in-time (JIT) module. The FP8GEMM library supports V3/R1 training and inference.

ChainCatcher news, sources say that DeepSeek is accelerating the launch of its R2 artificial intelligence model, which was initially planned for release in May, but is now working hard to launch it as soon as possible.

ChainCatcher news, DeepSeek today launched "Open Source Week," with the first open-source code repository being Flash MLA------an efficient MLA decoding kernel optimized for Hopper GPU, specifically designed for processing variable-length sequences.According to reports, Flash MLA is inspired by FlashAttention 2&3 and the cutlass project.

ChainCatcher news, the "2025 HTX DAO Victoria Harbour Night - Confidence Journey Hong Kong Station" hosted by HTX DAO was grandly held. Sun Yuchen, Global Advisor of Huobi HTX and Founder of TRON, revealed on-site that Huobi HTX has made significant progress over the past three years relying on the community, especially excelling in trading volume and new asset listings. Additionally, Huobi HTX is experiencing a surge in traffic in the CIS region in 2024 and will strongly focus on laying out the crypto market in the CIS region, emphasizing the resolution of user needs in that area to solidify its market position.Moreover, during the "Hong Kong TRON & HTX DAO Exchange Meeting," Sun Yuchen discussed the decentralized stablecoin USDD, stating that various methods have been tried to enhance user returns and improve user experience. USDD focuses on long-term development, with the key being to find the right team and leaders, possessing solid underlying technology and a sound community governance mechanism. During the construction process, it is essential to ensure substantial progress and prevent sudden project collapses due to issues like asset security, as only in this way can sustainable and healthy growth be achieved.Sun Yuchen also revealed that the application potential of AI in the crypto field is enormous, especially in building expert models. Currently, there is an exploration of an AI application based on DeepSeek with a certain level of anonymity, tailored to the characteristics of the crypto field.

ChainCatcher news, the Pando Fund released a research report titled "Decentralized AI Changed by Deepseek," which points out that the narrative of decentralized AI is continuously reshaping. Web3 AI companies can focus on replicating DeepSeek's success while offering new advantages such as multimodality, user ownership, censorship resistance, and privacy. It is expected that the number of projects on the supply side will continue to grow, while consumer-facing projects will begin to compete with Web2 counterparts by building community-participated networks. In the coming year, AI models developed based on a complete Web3 AI stack will emerge one after another.In addition, companies combining AI and crypto are gradually adjusting their strategies to focus on infrastructure building rather than model development. For example, companies in the GPU market like Akash, Render, IoNet, and Exabits have developed sustainable revenue models. Meanwhile, businesses like Grass and Gradient, which allow users to share network bandwidth, have found their market positioning by providing distributed network services to Web2 clients. In inference tasks, the performance gap between small models and large models is narrowing, which means that Web3 does not need to rely on the massive computing power of traditional AI giants and can utilize these streamlined models for efficient inference operations. As this trend develops, we may see more inference endpoints driven by decentralized computing networks in the future.

ChainCatcher message, the latest data from the data analysis platform QuestMobile shows that since its launch until February 9, the cumulative downloads of DeepSeekApp have exceeded 110 million, with a peak of nearly 97 million weekly active users.

ChainCatcher news, according to Jinshi reports, Baidu Search has fully launched the DeepSeek full version and is providing online services. Within just 1 hour of opening on the PC side, over ten million users have already used it.Compared to other products that integrate DeepSeek, users can experience it by clicking "AI Search DeepSeek Full Version" below the search box on the Baidu Search homepage. It is worth noting that Baidu Search has integrated the DeepSeek full version this time, which combines Baidu's online search capabilities and features enhanced retrieval technologies like RAG, making it easier for users to obtain updated and less misleading information.

ChainCatcher news, according to Jinshi reports, Weibo Zhiso has integrated DeepSeek and is currently undergoing internal testing. This integration of DeepSeek by Weibo will not only help improve the quality and experience of Weibo search but also provide users with intelligent services in more scenarios.