FinOps for AI has moved from a niche discipline to the single biggest cost-control question facing technology leaders, and most organisations are discovering it the hard way — through an invoice nobody forecast. A large language model that costs pennies per request in a pilot becomes a six-figure annual line item once it is embedded […]
SpaceX completed its $60 billion all-stock purchase of Anysphere, the maker of Cursor, on 14 August 2026 — the largest acquisition of a venture-backed startup ever recorded. Here is exactly what closed, how a four-year-old company reached that valuation, why SpaceXAI wanted it, what happens to the Cursor brand and its third-party models, and the vendor-concentration and data-governance questions your engineering team should work through in the next 90 days.
Anthropic turned the Claude in Chrome side panel into a full Cowork session on 12 August 2026, weeks after putting a sandboxed browser inside Claude Code on the desktop. Here is what has actually shipped, how a dedicated Cowork browser compares with the extension, which capabilities each surface gets, what Anthropic has built to contain prompt injection, and the governance your team needs before switching any of it on.
With Gemini users now able to disable the visible watermark on AI images, videos and songs, the invisible SynthID layer is what actually identifies AI content. Here is how SynthID hides a signal in pixels, audio and text, the 100-billion-file scale it has reached, every tool that can check a file for it, what it survives, and the verification workflow your team should adopt.
Twitch quietly enrolled every channel in Amazon AI training and shipped the opt-out afterwards — its CPO admitting “if it was opt-in, nobody would opt in.” Here is exactly what the Training for Generative AI setting covers, the four-click opt-out, the loopholes it leaves open, and our take on what consent-by-default means for creators and brands.
Cory Doctorow’s The Reverse Centaur’s Guide to Life After AI promises a route out of the AI hype cycle, but a sharp review in The Conversation calls it an outdated map that fails to answer its own question. Our opinion: both sides hold half the truth, and the argument matters for any business planning what comes after the bubble.
OpenAI shipped three ChatGPT upgrades in five days: interactive quiz generation, restaurant reservation search through OpenTable, Resy and Yelp, and a Google Drive integration that can browse and even edit Docs, Sheets and Slides from the chat. Here is what each feature does, which plans get it, and what the new write access means for businesses.
Apple has trained its own AI model for the Chinese market with Alibaba’s support, becoming the first foreign firm approved by Beijing to offer a proprietary generative AI model. Here is what the new Apple AI model does, how Alibaba Qwen and Baidu fit around it, and when Apple Intelligence finally launches in China.
Google will now allow users to remove the visible watermark from its AI generations. Here is how the new Gemini watermark toggle works across Nano Banana, Omni and Lyria, what SynthID and C2PA metadata still record, who gets the option, and the risks for trust online.
AI research agents can read the literature, write analysis code and draft convincing papers — yet benchmarks show they fail most realistic research tasks and cannot yet originate science. Here is the evidence, why the gap persists, and how to use them safely.
