The top things worth knowing about in AI today.
Apple released iOS 27 on Monday with a rebuilt Siri running on a mix of its own foundation models and Google Gemini, under a deal reported at roughly US$1 billion a year. The assistant holds context across a back-and-forth, runs as a standalone app with chat history, and is opt-in on iPhone 15 Pro and later, English first with more languages from October and no EU availability at launch. It is the largest consumer AI rollout yet to run on a rival lab model.
Read moreMeta launched Meta One, a paid tier for its AI tools across Instagram, WhatsApp, Facebook and Messenger. Consumer plans run from US$2.99 to $19.99 a month and creator and business plans from $14.99 to $499, with the company reporting 15 million subscriptions and trials so far. Basic Meta AI stays free, so the question is whether enough people pay for the extras to offset Meta capital spending.
Read moreAnthropic released Claude for Financial Advisors, connecting Claude to Charles Schwab, BlackRock, Vanguard, Addepar, Envestnet, Orion and iCapital, alongside eight prebuilt workflows covering pre-meeting prep, portfolio rebalance review, estate and tax briefs and compliance checks. Claude will not give investment advice, and every client-facing action needs adviser approval. It is a clear shift toward selling regulated-industry workflows rather than a general chatbot.
Read moreNvidia-backed Firmus has begun a roadshow for an ASX listing of up to A$7 billion, targeted for late October, which would rank as Australia second-largest IPO behind the 1997 Telstra privatisation. The raise funds Project Southgate, a 1.6GW network of liquid-cooled AI data centres anchored in hydro-powered Tasmania with sites planned for Melbourne, Sydney, Canberra and Perth. Australian institutions are about to price AI infrastructure risk directly.
Read moreOpenAI has acquired Glass Imaging for more than US$300 million according to the Wall Street Journal, about triple the startup $100 million valuation last year. Glass was founded by two former Apple engineers who worked on Portrait Mode, and uses neural networks tuned to a specific camera optics to improve images as they are captured rather than edited afterwards. It adds computational photography staff to OpenAI consumer hardware effort.
Read more404 Media obtained internal documents describing Project Lily, under which OpenAI pays hundreds of contractors, recruited through Crossing Hurdles and paid via Mercor with some earning above US$50 an hour, to read real ChatGPT conversations and score responses on a one to seven scale. Usernames are stripped, but reviewers can see full conversations and a summary of the user stored memories. A useful reminder for staff about what goes into a consumer chatbot.
Read moreAnthropic published engineering data showing Claude now authors about 80% of the code merged into its production codebase, with engineers shipping roughly eight times more code per quarter than they did between 2021 and 2025. Tests across the codebase grew tenfold and continuous integration jobs rose 25-fold in six months, forcing three rewrites of its test-selection service. Teams adopting coding agents should expect the bottleneck to move from writing code to reviewing and testing it.
Read moreChina foreign ministry responded to Anthropic chief executive Dario Amodei call for tighter chip export controls and a crackdown on distillation, with spokesperson Guo Jiakun saying fearmongering, confrontation and vicious competition would only disrupt global AI governance. State outlet Global Times accused him of waging a silent AI Cold War. The exchange makes coordinated international rules on frontier AI less likely in the near term.
Read moreA short read every week — the few things worth your time, and nothing that isn’t.