Group 1 - The article titled "2028 Global Intelligence Crisis" predicts a future where the unemployment rate in the U.S. could soar to 10.2% and the S&P 500 could drop by 38% from its peak, suggesting that as AI becomes more advanced, economic conditions worsen [1][2] - A negative feedback loop is proposed where increased AI capabilities lead to more layoffs, which in turn allows companies to invest in more AI tools, further enhancing AI capabilities and resulting in more job losses [1][3] - The article highlights that AI could dismantle traditional business models and industries, leading to a potential collapse of the $13 trillion mortgage market as even high-credit borrowers may struggle to meet loan obligations after being replaced by AI [1][3] Group 2 - Several Wall Street institutions, including Castle Securities and Deutsche Bank, have criticized the article's extreme assumptions, arguing that the savings from layoffs cannot sustain continuous AI advancements and that the premise of "frictionless" AI adoption is unrealistic [2] - Critics assert that the report lacks data support and resembles an emotionally driven apocalyptic narrative rather than a grounded analysis [2][3] - The core issue raised by the article touches on the economic system's reliance on the assumption that human intelligence is a scarce resource, which may be challenged by rapid AI development [3] Group 3 - The article concludes by emphasizing that the negative feedback loop has not yet begun and that there is still time to adapt before 2028 [4] - Recommendations for the future include critically assessing income sources based on the assumption of being better than machines, leveraging AI to become decision-makers rather than easily replaceable workers, and recognizing the revaluation of "human traits" that cannot be easily replicated by AI [5][6][7]

Group 1 - AMD has secured a significant partnership with Meta, involving a collaboration worth over $60 billion for deploying AMD Instinct GPUs, with an overall scale estimated to exceed $100 billion [1] - The core of the collaboration is a customized GPU based on the MI450 architecture, adhering to the "Workload First" principle, with shipments expected to begin in the second half of 2026 [1] - Meta's infrastructure head stated that a single chip cannot meet all workloads, indicating a trend among leading AI players to diversify computing power to mitigate supply chain risks [1] Group 2 - Anthropic has introduced a Remote Control feature for Claude Code, allowing users to remotely connect to local projects via browser or mobile, enabling real-time monitoring and control [2] - The Remote Control supports two initiation methods: creating a new remote session or integrating into an existing conversation, with connections established through QR codes, URLs, or a code list [2] - Unlike Claude Code on the Web, which runs on cloud virtual machines, the Remote Control executes code locally while the mobile device acts solely as a remote control [2] Group 3 - Anthropic has updated its Claude Cowork plugin system, enabling users to customize AI plugins from scratch through conversational guidance, with initial templates available for various verticals [3] - The plugins are deeply integrated with enterprise tools like Slack, Salesforce, and Excel, allowing for cross-application context continuity, and enterprise admins can create private plugin markets [3] - The addition of OpenTelemetry support allows for quantifiable AI input-output, as Anthropic transitions from a tool provider to a platform, establishing a foundation for enterprise AI infrastructure [3] Group 4 - xAI's Grok image-to-video model achieved the top ranking in the Image-to-Video Arena with an ELO score of 1404, surpassing 34 other models based on 465,000 blind test votes [4] - Grok Imagine 1.0 can generate 10-second 720p videos with native audio and offers capabilities for text-to-video, image animation, and zero-threshold video editing, with an API pricing of approximately $4.20 per minute [4] - The model excels in instruction adherence, cinematic shot control, and lip-syncing, while also leading in quality, latency, and cost balance, supporting multi-turn interactive creation [4] Group 5 - Alibaba has open-sourced three medium-scale models under the Qwen 3.5 series, with the 35B-A3B model surpassing its predecessor, the larger Qwen3-235B-A22B [6] - These models utilize a mixed attention mechanism and high-sparsity MoE architecture, achieving state-of-the-art results in various authoritative benchmarks [6] - The Qwen 3.5-27B model is the first dense model to exceed GPT-5 mini in tool invocation and programming, and it can run on a single GPU, with the Flash version API costing only 0.2 yuan per million tokens [6] Group 6 - MiniMax has launched the MaxClaw mode on its Agent platform, allowing for one-click configuration of OpenClaw, with pre-set tools and skills that can be activated in 20 seconds [7] - The Expert community has accumulated over 10,000 public expert agents across various fields, enabling users to create custom agents through natural language dialogue without coding [7] - Future plans include establishing a Marketplace for users to list and price their self-built experts [7] Group 7 - A Cloudflare engineer rebuilt the Next.js framework in a week using AI, creating a replacement called vinext, which was built on Vite and involved approximately 800 AI sessions costing $1,100 [8] - The client bundle size is reduced by about 57% compared to Next.js, and the new framework has passed over 1,700 unit tests and 380 end-to-end tests, covering 94% of Next.js APIs [8] - This case demonstrates that AI can lead large system implementations under clear architectural specifications and mature foundational tools, redefining many abstract layers in software [8] Group 8 - The U.S. Department of Defense has pressured Anthropic to lift AI safety restrictions on Claude for military secret systems, threatening to invoke the Defense Production Act if demands are not met [9] - xAI's Grok has accepted military conditions to enter classified systems, with Google and OpenAI also in discussions, creating competitive pressure on Anthropic [9] - Anthropic has released RSP 3.0, abandoning its previous commitment to a unilateral training pause, shifting from "absolute risk" to "marginal risk" assessment [9] Group 9 - CitriniResearch's report titled "2028 Global Intelligence Crisis" has garnered over 10 million reads, predicting a negative feedback loop where AI capability increases lead to layoffs and reduced consumer spending [10] - The report warns that white-collar workers make up half of U.S. employment and drive three-quarters of discretionary spending, with AI potentially disrupting various platforms [10] - The risk extends to a $13 trillion mortgage market, with San Francisco home prices dropping 11% year-on-year, highlighting systemic risks due to the accelerating pace of AI capabilities compared to institutional adaptation [11]