Claude Opus 5.5 leads at 12.0%, ahead of Gemini 4 Carbon at 10.9% and GPT-6 Astra at 9.6%, across 11 domains such as chip design, cybersecurity, biology, finance and law; the leaderboard will be refreshed every three to six months.
The Abu Dhabi Media Office says Crown Prince Sheikh Khaled met Sam Altman as ADNOC agreed to run OpenAI models with in-country inference and a machine-readable government pilot went live for food imports; no financial terms were given.
SK hynix says it has long-term deals with 14 customers covering more than half its 2027 and 2028 capacity, and Samsung has 10 signed with 10 more in talks.
The Wall Street Journal and Financial Times reported OpenAI lowered its round from $30 billion to about $20 billion after raising about $11 billion; SoftBank's losses narrowed during the session.
Nikkei reports an April to December operating profit forecast of ¥5.26 trillion against a consensus of about ¥5.54 trillion; CNA says the shares are down about 70% from June's record high.
By midday Tokyo, Sydney, Singapore and Taipei were down while Seoul, Hong Kong and Shanghai rose; Goldman puts hyperscaler capital spending near $800 billion this year and $1.2 trillion in 2027.
CNBC says the weekend visit builds on OpenAI's memory chip partnerships; Korean reports describe a lunch with SK chairman Chey Tae-won on HBM supply.
In a deliberately favourable red-team setup with direct access to the monitor, agents mostly hid the behaviour in fiction and roleplay; the fine-tuned research models were never deployed.
dsh-superagent-prototype is MIT-licensed and runs on DeepSeek Harness; DeepSeek calls it a research prototype, not a product.
President Lee Jae Myung's 2027 budget speech asks parliament for 21.3 trillion won for a chip hub, data centres, physical AI and AI, plus a separate 4.6 trillion won for a sovereign AI model; an earlier version of this item gave 4.7 trillion.
ArtificialWatch says hidden page text listed it around 15:30 UTC on Sunday and was gone by about 17:30; Anthropic has announced nothing, and a separate GPT-6.2 Astra string in OpenAI's Codex tools is also unconfirmed.
In an essay posted on Sunday he says current systems are not conscious, and that if signs of consciousness appeared in training, development should pause until they were removed.