Meta's new AI agent, Muse, has leapt to the top of app charts roughly two weeks after its release. Its popularity has helped push up chip stocks by focusing attention on the importance of CPUs for agentic AI.
Taiwan's Legislative Yuan passed the Virtual Asset Service Act in June 2026, and Taiwan's Financial Supervisory Commission (FSC) is now drafting delegated regulations, with completion planned for the first quarter of 2027. PwC Taiwan managing director Monica Lee said banks will need licenses to operate virtual asset services, and lending licenses are the most likely to be profitable.
DeepSeek CEO Liang Wenfeng said during a call with investors that he aims to rely on Huawei chips to train the startup's models in the future, and that a batch of training chips from Huawei are set to begin arriving at his company as early as the fourth quarter of 2026.
Chinese labs have spent the year arguing that reinforcement learning, not pre-training, holds the next gains. Xiaomi has now put a price on that argument. The company released and open-sourced its MiMo-V2.6 series on September 22, after livestreaming the reinforcement learning post-training run that produced it — publishing step counts, token consumption, and a running cost meter on a public page as the models trained.
As artificial intelligence advances toward greater autonomy, balancing capabilities with safety has become a central challenge for industry leaders. Emerging model behaviors and high-profile security incidents have prompted a continuous evolution of safety initiatives, internal engineering adjustments, and proposed joint testing efforts between major AI developers.
Google has opened preorders for the first Googlebook models, with prices starting at US$899, as it pushes Gemini AI deeper into the notebook experience and links phones, apps, and desktop workflows through a new operating system that blends Android and ChromeOS elements. The launch targets Android users, developers, and the premium notebook market.
