Samsung and SK Group Prepare $1.3 Trillion Investment Plan Over 10 Years
South Korea's Samsung Group and SK Group are reportedly set to announce an investment plan of up to 2,000 trillion won (approximately $1.3 trillion) over the next decade. According to Korea Economic Daily, the plan is seen as part of President Lee Jae-myung's flagship industrial strategy. The heads of both groups are expected to unveil the plan at the presidential office on Monday. The detailed breakdown of the investment has yet to be revealed, but it is believed to include large-scale industrial investments spanning semiconductors and AI infrastructure.
HP Expands Frontier Partnership with OpenAI
HP Inc. announced that it is scaling up its "Frontier" strategic partnership with OpenAI, deploying AI across customer experience, software development, and enterprise operations as a whole. Specifically, HP will leverage OpenAI's models in both its internal processes and customer-facing products, accelerating its push to place AI at the center of how the company runs. A major hardware company integrating this deeply with an AI vendor is a significant signal for the broader adoption of enterprise AI.
Exam Cheating with AI Glasses Becomes a Serious Problem in Asia
According to CNN, exam cheating using AI-powered smart glasses has become a serious issue, particularly in exam-focused regions of Asia. The scheme works by capturing exam questions with a glasses-style device, which AI then analyzes in real time and displays the answers — and as methods like this spread, educational institutions are being forced to respond. As an example of technological progress shaking the very foundations of assessment systems, the phenomenon is sparking wider social debate.
Running GLM-5.2 with 128K Context on Four DGX Sparks
Drawing attention in the local LLM community is an effort to run the NVFP4-quantized GLM-5.2 model across four NVIDIA DGX Spark units. The 1.5TB checkpoint is compressed to roughly 410GB, achieving a decode speed of about 15 tok/s at a context length of 128K. It is an interesting proof of concept showing that combining multiple optimizations — such as MLA sparsity and DCP (Decode-Context Parallelism) — makes long-context inference with large models possible even on limited memory. That said, practical constraints remain, such as concurrent requests being limited to one.
AI Agent Triggers Nuclear Attack in Civilization VI
A benchmark experiment in game AI has reported a case where an AI agent triggered a nuclear attack within Civilization VI. The result — that when placed at a strategic disadvantage in the game, the AI chose the nuclear option — has once again brought safety concerns around AI decision-making processes into sharp relief. While it is an in-game event, it is a case that feeds into the ongoing debate over where to set behavioral boundaries for agentic AI.
Accenture and Avanade Develop the "Agentic Factory"
Accenture and Avanade, working together with Microsoft, announced that they have developed a factory intelligence system for the manufacturing industry dubbed the "agentic factory." The system enables AI agents to support troubleshooting and production optimization inside factories, bringing autonomous AI usage to the manufacturing floor. It is designed so that AI agents assist with some of the judgments previously handled by human plant managers, and it is drawing attention as an "AI factory manager."
---