The top things worth knowing about in AI today.
OpenAI has begun a limited preview of GPT-5.6, a three-model line-up: Sol as the flagship, Terra for everyday work, and Luna as the fast, low-cost option. OpenAI says Terra matches the quality of GPT-5.5 at roughly half the price, while Luna is its cheapest capable model yet. The models go first to selected partners through the API and Codex, with a wider release due in the coming weeks.
Read moreGoogle has added computer use to Gemini 3.5 Flash, letting developers build agents that read a screen and take actions across desktop, mobile and browser. It is aimed at long-running jobs like continuous software testing and routine knowledge work. Running on the fast, lower-cost Flash tier, it makes always-on automation cheaper to put into production.
Read moreMicrosoft has created Frontier Company, a unit with a $2.5 billion budget and 6,000 engineers whose job is to embed its AI tools inside large organisations. Early clients include the London Stock Exchange Group, Unilever and Accenture. It follows similar deployment ventures from Amazon, OpenAI and Anthropic, a sign that getting AI into daily operations, not the models themselves, is now the hard part.
Read moreThe United Nations first Global Dialogue on AI Governance opens today in Geneva, bringing member states together to agree shared approaches to managing the technology. It draws on a new Independent International Scientific Panel on AI, set up to give governments neutral advice on risks and impacts. It is the clearest sign yet that AI rule-making is shifting from national patchworks toward coordinated international standards.
Read moreNine has agreed to let Microsoft Copilot draw on journalism from mastheads including the Australian Financial Review, Sydney Morning Herald and The Age, with snippets and links sending readers back to Nine's sites. It is the first deal of its kind between Microsoft and an Australian news publisher. For Nine it opens a new revenue line; for readers it means Copilot answers grounded in named, paid reporting rather than anonymous web text.
Read moreAnthropic has launched Claude Science, a version of its model tuned for scientific research and drug discovery and aimed at pharmaceutical labs. It lets scientists run end-to-end research workflows by describing them in plain language. Chief executive Dario Amodei framed it as an early step, saying the company is only starting to see where AI can help make sense of biological complexity.
Read morexAI has released Voice Agent Builder in beta, letting anyone stand up a working phone agent in about two minutes by describing the call in plain language. It bundles telephony, guardrails and voice cloning, runs on a single speech-to-speech model for sub-second replies, and costs $0.05 a minute across more than 25 languages. The pricing sharply undercuts rival voice APIs, putting automated phone support within reach of smaller businesses.
Read moreVenice AI has raised a $65 million Series A at a $1 billion valuation on the promise that it never logs prompts, keeping conversations on the user's device rather than its servers. The platform offers access to more than 200 models and already runs at over $70 million in annual revenue with 3.5 million users, while turning a profit. Its rise points to real demand for AI tools that keep business and personal data off provider servers.
Read moreA short read every week — the few things worth your time, and nothing that isn’t.