This was a big week for the tools you actually use at your desk. OpenAI launched GPT-5.6 and a version of ChatGPT that can stick with a project for hours and get it done. Meanwhile Anthropic made a quiet governance move that says a lot about who it wants steering AI.
OPENAI
GPT-5.6 arrives, and ChatGPT starts doing the work
OpenAI released a new model family led by GPT-5.6. The pitch is simple: more useful output from every dollar you spend, with gains in tricky areas like cybersecurity. It is also now the default engine inside Microsoft 365 Copilot, so it will quietly power Word, Excel, PowerPoint, and Chat for millions of people.
The bigger shift is ChatGPT Work, an agent that takes action across your apps and files. Instead of just answering a question, it can take a goal and stay on it for hours until the job is finished.
The news lands during a rocky stretch. There is open chatter about the OpenAI and Microsoft partnership straining, and OpenAI's No. 2 executive, Fidji Simo, just stepped down after a longer-than-expected medical leave.
it can take a goal and stay on it for hours until the job is finished
Why it matters The AI in your everyday office apps just got smarter and can now finish multi-step tasks, not just chat about them.
ANTHROPIC
Anthropic adds Ben Bernanke to its trust and asks for hard questions
Anthropic appointed Ben Bernanke, the former US Federal Reserve chair, to its Long-Term Benefit Trust. That trust is the unusual body that helps hold the company accountable to its mission rather than just to profit. Bringing in a respected economic steward signals Anthropic wants serious outside judgment on big decisions.
The company also opened a public call for the hardest questions people have about AI, and promised to show its work as it answers them.
Separately, Anthropic is teaming with UST to bring Claude into physical AI, meaning robots and real-world machines, not just software.
opened a public call for the hardest questions people have about AI
Why it matters Who governs these companies shapes how safe and trustworthy the AI you build on will be.
AGENTS
A startup let its own AI agent run a $100M raise
Lyzr, a company that builds AI agents for businesses, did something bold: it let one of its own agents run its $100 million fundraise. The move doubled as a live demo. If the product can handle something as high-stakes as raising money, the thinking goes, it can handle your workflow too.
It is a sign of where things are heading. Agents are moving from helping with tasks to owning whole processes end to end.
it let one of its own agents run its $100 million fundraise
Why it matters Agents handling real, high-stakes business processes is no longer a demo idea, it is starting to happen.
BROWSERS
OpenAI shuts down Atlas, keeps the browsing ambitions
OpenAI is sunsetting Atlas, its AI-powered web browser, after less than a year. But it is not giving up on the idea of AI that browses the web for you.
Instead, the useful parts are moving into its desktop app and a Chrome extension. The lesson for builders: even big companies kill products fast and fold the good bits into where users already are.
even big companies kill products fast and fold the good bits into where users already are
Why it matters It shows that betting your workflow on a brand-new AI app is risky, since even OpenAI retires them quickly.
HEALTHCARE
An AI blood test could spare women invasive cancer checks
Several NHS hospitals in England are preparing to use an AI-powered blood test to assess women referred for possible womb cancer. Around 90,000 postmenopausal women are referred each year for checks that are often uncomfortable and invasive.
The AI test aims to sort out who truly needs the invasive procedure and who does not. That means less discomfort for many patients and faster answers.
That means less discomfort for many patients and faster answers.
Why it matters It is a concrete example of AI making a real health service faster and kinder, not just cheaper.
Which of these would actually change your workweek? Hit reply and tell me, I read every response and often turn them into future issues.
Free · every Friday
Get the weekly AI brief
Plain-English AI, in your inbox each week. One email, no spam.
No spam, ever. Unsubscribe with one click. Phone is optional and only used if you ask us to call.
Prefer the website? Subscribe at openhour.io →
