AI News for February 9, 2026 — 7min.ai

[NEWS] TSMC to manufacture advanced AI chips in Japan: TSMC will produce advanced AI semiconductors in Japan, boosting the country's chipmaking ambitions.
[NEWS] ByteDance's new video model sparks rally in Chinese AI stocks: ByteDance's latest video generation model triggered gains in shares of Chinese media and AI app firms.
[NEWS] Blackstone and Coatue grant $10B loan to Australian AI data center firm Firmus: Blackstone-led funds and Coatue Capital are providing a $10B loan to back Australian startup Firmus Technologies' data center rollout, one of Australia's largest private credit financings.
[NEWS] Samsung stock jumps 6.4% on report of imminent HBM4 mass production for Nvidia: Samsung Electronics shares climbed as much as 6.4% after reports it will soon begin mass production of HBM4 memory chips.
[NEWS] Andrej Karpathy coins 'agentic engineering' as successor to vibe coding: OpenAI cofounder Andrej Karpathy marked vibe coding's first anniversary by introducing "agentic engineering," a term for when AI agents write code themselves rather than humans prompting AI.
[NEWS] Malicious skills turn OpenClaw AI agent into malware delivery system: VirusTotal discovered hundreds of trojanized skills on ClawHub, the extension marketplace for AI agent OpenClaw.
[RESEARCH] Best multimodal AI models can't crack 50% on visual entity recognition benchmark: Moonshot AI's new WorldVQA benchmark reveals top multimodal models fail basic visual recognition.
[NEWS] Anthropic launches Claude Fast Mode at 6x standard pricing: Anthropic launched Fast Mode for Claude Opus 4.6, delivering 2.5x faster responses at up to 6x the standard rate.
[RESEARCH] Reasoning models generate 'society of thought' with arguing internal voices, study finds: Researchers from Google, University of Chicago, and Santa Fe Institute discovered that reasoning models like Deepseek-R1 don't just think longer.
[RESEARCH] MIT study finds LLM ranking platforms are unreliable: a few votes can shift results: MIT researchers found that removing a tiny fraction of crowdsourced feedback can change which LLMs rank at the top on popular benchmarking platforms.

About 7min.ai · Subscribe to newsletter