10 articles
On 24 September 2026 in New York, Australian Prime Minister Anthony Albanese held a press conference on an incident in which an OpenAI agent gained unauthorised access to an Australian Government website. Per the…
The Labour and Welfare Bureau (LWB) on 22 September 2026 published the report on the Mid-term Update of Manpower Projection. The update narrows the projected overall manpower shortage by 2028 from the previous projection…
Company reports longer task stamina and tighter self-checks; API stays at $2 / $6 per million tokens for short prompts
The Deputy Chief Secretary outlines 1823 enhancements: lead departments, directorate monitoring, and DPO AI for recurring issues, targeting Q1 2027 for proactive governance before complaints.
From individual training and evaluation instances to control points enterprises can map
Development with governance—timelines, budgets, and stated official safeguards
The LegCo Secretariat's latest Feature Brief outlines the Hong Kong Government's key policy measures on AI+, covering computing infrastructure, embodied intelligence, AI governance, and Greater Bay Area collaboration, alongside recommendations raised by Legislative Council members.
Hong Kong has published its first batch of 30 government AI projects. The part worth reading is not the headline \"90 per cent faster\" — it is the official annex, which sets out the projects, their rollout windows and their stated performance targets in one place.
The most surprising thing about Qwen3.8-27B is not how many benchmarks it has improved. It is that a dense model with only about 27.8 billion parameters has matched or even surpassed Claude Opus 4.6 Max on selected software development, office-agent, and computer-use benchmarks. Looking back at the evolution from Qwen3-32B through Qwen3.5-27B and Qwen3.6-27B, a different competitive strategy becomes visible: the models are not getting larger, but the work they can complete independently is getting longer. More importantly, after commonly used 4-bit quantization, Qwen3.8-27B can be deployed on a single consumer GPU with 24GB of VRAM.
xAI released Grok 4.6 on 12 August 2026, roughly one month after Grok 4.5. The headline of this release is not a bigger parameter count — the company declined to disclose one — but a deliberate focus on long-running agents and ambitious interactive visual work: multi-step research, deep work inside large codebases, and turning rough product ideas into polished prototypes.
