This week: AI agents that work on their own, Google's Gemini laptops go on sale, and scam texts get harder to spot

A Friday read on what actually happened across AI, hardware, and the web. Curated, not chased. If nothing mattered, we’d say so.

AI agents that don’t wait to be asked

At its developer conference on 29 September, OpenAI introduced Dots — a step on from a chatbot that waits for your next message. A Dot stays active and works on a task on its own: it gets its own cloud computer and browser, and can act across ChatGPT, email, Slack and text — checking what came in overnight, chasing jobs to completion, editing documents, rearranging your calendar when a meeting moves. For now they’re on OpenAI’s paid Pro and Business plans, not free accounts.

Why it matters to you: an assistant that keeps working after you’ve closed the chat is genuinely useful, and it is also doing things in your name while you’re not watching. The setting that matters is the permission: OpenAI lets you require a Dot to ask before it does anything that counts — sending a message, spending money, changing a file. Switch that on. The convenience is new; what hasn’t changed is that you’re accountable for what an agent does on your behalf, so keep a hand on the actions that are hard to undo. The same week, a president tried to rename AI itself — our editor on why the label you’re handed is a power move: The name is a power move.

Verify it yourself: 9to5Google — OpenAI launches Dots, ‘always-on agents’ you can assign tasks to · CNBC — OpenAI DevDay 2026

Google’s Gemini laptops go on sale

Google’s first Googlebooks reach shops on 4 October in the US and 5 October in the UK. They’re a new kind of laptop built around Google’s Gemini AI and designed to work hand-in-glove with an Android phone — open your phone’s apps and files straight from the laptop, with Gemini suggestions built into the cursor and desktop. Models come from Acer, ASUS, Dell, HP and Lenovo, starting at $899 in the US (UK prices via Google) — premium ground, not the cheap-Chromebook bracket Google is known for.

Why it matters to you: this is the biggest change to Google’s laptops in years, and the first wave is aimed squarely at people already living in Android and Google’s apps who want AI baked into everything. If that’s you, it’s worth a look — but a brand-new platform is the kind of thing to let other people test first. Wait for real-world reviews, weigh the price against a well-specced Windows or Mac machine, and be sure you actually want the AI-and-Android angle before paying the premium for it.

Verify it yourself: Google — Introducing Googlebook · 9to5Google — Googlebook pre-orders start 21 September

Scam texts are getting harder to spot

The UK’s National Cyber Security Centre (NCSC) and the consumer group Which? have both flagged how convincing scam messages have become. Fake “missed parcel” texts posing as Royal Mail, along with bank and HMRC impersonations, now read cleanly — no clumsy English, no obvious typos. AI writing tools are part of why, which means the old advice to “spot the bad spelling” no longer protects you.

Why it matters to you: if you can’t reliably catch a fake by looking, stop trying to and verify instead. Never tap a link in a text or email you weren’t expecting; if it claims to be your bank, a courier or the taxman, go to them directly through their app or a number you already have. Then report it: forward a scam text to 7726 (free on most UK networks) and send a scam email to report@phishing.gov.uk. It takes seconds and it works — more than 40,000 scams reported this way have already been taken down. For the step-by-step, see our guide to checking a link before you click.

Verify it yourself: NCSC — how to spot and report scam texts · Which? — the most convincing scams of 2026 so far

Spotted something we should cover next week? That’s the whole idea — tell us, and we’ll check it.