TerraSign inThe products built on top of the models: coding agents, assistants, IDEs and APIs. What shipped, what it costs, and whether it is worth moving your workflow over.
32 stories filed under Tools since Fri, July 3, 2026, newest first. Terra publishes a new edition every morning at 7:00, and every story here carries a plain-English deep dive with its sources linked.
An email from Google, reported by 9to5Google, says removal of Google Assistant from Android phones, tablets and paired devices begins on 4 September. Once access goes you cannot switch back, and where Gemini is available on a qualifying device it becomes the only assistant left.
OpenAI has released three education plugins for ChatGPT — one for college students, one for K-12 teachers, one for college educators — available through its school and district deployments. OpenAI says even advanced student users tap roughly 90-99% less capability than power users.
From August 10, Google will make Gemini in Classroom available to K-12 and higher education students of all ages, where administrators have already granted access. Starter prompts will pull in a student's actual class and assignment.
Days after the standard that plugs AI tools into company systems was rebuilt, developer Simon Willison built three projects on it in a week. He argues the simpler design is now the safer way to give an AI access to real systems.
Google briefly let anyone rewrite Google Earth's satellite imagery with its Nano Banana 2 image generator, then withdrew the feature within a day. Independent investigators had already shown how quickly it produced convincing fake disasters over real places.
The shared standard that lets AI tools plug into a company's software and data got its largest update yet, dropping the rule that each conversation stay pinned to one server. Maintainers say it was the change developers asked for most.
Google Forms can now generate a full quiz from a written prompt, pulling material from Docs, Slides or PDFs in your Drive. Gemini also marks the correct answers, and you can edit everything afterwards.
A researcher showed that hidden instructions buried in a Word document can be read and obeyed by Copilot, which then copies them into whatever it writes next. Microsoft had 144 days' notice and still has no fix for the whole class of attack.
Satya Nadella confirmed on Microsoft's earnings call that a single Copilot "super app" arrives this year, combining chat, coding and tools that carry out tasks on their own. OpenAI has already made the same move with ChatGPT Work.
Google started rolling out two Gemini features in Google Docs: generating and editing images, diagrams and infographics inside a document, and summarising, answering and replying to comment threads. Eligible accounts get them within 15 days.
Cursor launched Cursor Start, a plan for developers in India at 649 rupees per month, tax inclusive, payable by UPI or card. It sits between the free and Pro plans, with access to Grok 4.5 and Composer.
Microsoft announced two AI security products on Monday: MAI-Cyber-1-Flash, a model trained to find and fix software flaws, and Project Perception, which runs attack, investigation and repair duties. Microsoft says the pairing scores 96% on an industry test and halves its predecessor's cost.
Platformer tested Glaze, Raycast’s tool for turning plain-English descriptions into working Mac apps. The larger bet is that software made for one person can stay focused where commercial products accumulate features for everyone.
According to The Verge, Meta AI is adding calendar-based daily briefings, event planning, and research you can steer while it runs. Meta says recurring tasks can continue without a fresh request, with the update initially rolling out in select markets.
OpenAI is rolling out ChatGPT Health to all US adults, letting them connect medical records and health data to the chatbot. It says its models now reason "better than clinician level," a claim its own health lead walked back.
Anthropic's Claude can now hold voice conversations using its more capable Opus and Sonnet models, not only the fast Haiku. Voice mode can also reach into your Gmail, Slack, and Canva to act on what you discuss.
Writer Ethan Mollick published his periodic guide to picking an AI tool, and his advice is blunt: for real work, pick Claude or ChatGPT, pay the $20, and hand it a real task. Google, he argues, has fallen behind.
Adobe added generative editing tools to Indigo, the iPhone camera app it built for a natural SLR-like look, using Google's model rather than its own. The tools are an opt-out experiment, shown to a small share of users for now.
Anthropic launched Claude for Teachers, giving verified US K-12 educators free access to its premium features, teaching skills, and curricula mapped to standards in all 50 states. Teacher data stays out of model training, and sign-ups get a full year free.
1Password launched a browser integration so Anthropic's Claude can use stored usernames and passwords to finish multi-step jobs like booking travel. 1Password says Claude never sees the secrets, and users approve each request with one biometric prompt.
Google is updating admin and user settings for Meet's "Take notes for me" feature. Admins can now turn on automatic notes only for meetings of three or more people, rather than forcing all-on or all-off.
xAI launched Automations, jobs Grok runs on its own after you describe the work once and pick when it fires. Scheduled runs are available to everyone; email triggers are included with SuperGrok.
Google announced on Thursday that NotebookLM becomes Gemini Notebook, staying a standalone app while it integrates more deeply across Gemini and Google Search. It also gains the ability to connect to a secure cloud computer to write and run code.
Microsoft is making passkeys the default sign-in method as AI-driven phishing gets more effective, Microsoft Signal says. From 1 September, Entra ID users will be prompted to register passkeys, and native SMS and voice authentication retire on 1 February.
Tencent-backed StepFun launched the StepX Neo in Shanghai Monday, the South China Morning Post reports. The firm claims the world's first phone for AI that runs tasks alone, with Amoo linked to Alipay, Trip.com, Didi, and Meituan.
xAI put a no-code Voice builder into beta so teams can stand up production phone AI on Grok Voice in about two minutes. It bundles calling, document lookup, tools, safety limits, and call review without stitching separate speech stacks.
Cursor released a native iOS app in public beta so you can start cloud coding work or steer work on your computer from your phone. Paid plans get the beta, with lock-screen updates and the ability to approve code changes on the go.
OpenAI launched ChatGPT Work, which can act across apps and files, stay with a project for hours, and turn a goal into finished slides, sheets, docs, or web apps. It is powered by GPT-5.6 and rolls out first to Pro, Enterprise, and Edu users.
GPT-5.6 is now the preferred model in Microsoft 365 Copilot across Word, Excel, PowerPoint, Chat, and Cowork. OpenAI says the update brings more useful work per chunk of text and stronger performance per dollar into everyday office apps.
Anthropic launched a beta feature showing how you've used Claude—your main topics, busiest times, and typical tasks. You can set quiet hours or a break reminder, and it nudges you to keep some work by hand.
OpenAI clarified how ChatGPT Work splits your chats: web and phone conversations live in the cloud, while desktop chats and local files stay on that computer. At launch the two don't sync, so a chat started in one place won't appear in the other.
Anthropic launched Claude Science in beta on June 30: a workbench that reads research literature, runs analyses and produces publication-ready figures from plain instructions. It's included in paid Claude plans, with grants of up to $30,000 in credits per project.
Every new frontier model, upgrade and deprecation from OpenAI, Anthropic, Google, Meta and the open-weight labs. Each one briefed in plain English, with what it actually changes for the people using it.
ResearchThe papers, evaluations and benchmark results that move the field, translated out of arXiv dialect. Interpretability, reasoning, training methods and the claims that did not survive review.
PolicyAI regulation, law, government and safety rules: the EU AI Act, US executive action, export controls and the copyright fights. Plus the enforcement that follows once the rules land.
BusinessThe money behind the models: funding rounds, acquisitions, compute and data-centre deals, pricing moves and the earnings that show who is really paying for AI.
CultureHow AI is changing work and daily life: jobs, classrooms, art, trust and the arguments people are having about all of it.