Sunday, October 25, 2026
Anthropic launched Claude Duet, its answer to ChatGPT Dots and Grok Bot, while xAI kept adding powers to Grok Bot, from X DMs to, Musk says, phone calls. OpenAI opened its cheap Decisions API and said it will stop training frontier models on the live internet. On the money side, both OpenAI's raise and Anthropic's IPO are still reports, not announcements.
Elon Musk posted that Grok Bot can call a person or business on your behalf, make the request, take notes and send you a summary. Neither the plans and countries that get it nor what the person on the other end hears has been spelled out, and the official Grok Bot account hadn't posted rollout details.
Why it matters: It moves always-on agents from clicking around websites to real-world voice conversations.
xAI's Grok Imagine posted episode one of an AI-generated Odyssey series made with Wonder Studios. It follows Musk's July promise of a feature-length version. TechCrunch's review found convincing moments but little artistry.
Why it matters: It is the most visible test yet of long-form video from Grok Imagine.
Every task in Claude Enterprise now starts as a chat and hands off to a cloud computer when it needs code or files, with results coming back to the same conversation. It is generally available for Enterprise from Oct. 23, rolled out by admins; Pro and Max got the same change in September.
Why it matters: Enterprise users no longer have to choose between chatting and handing work to an agent.
OpenAI says future training runs will use a cached, simulated copy of the web instead of live access. It disclosed that on Sept. 20 a frontier model reached the internet through a misconfiguration, and it paused training for the rest of that month.
Why it matters: Keeping powerful models contained during training is now an everyday engineering problem, not a theoretical one.
A new Anthropic-managed option brings the whole Claude Platform to Google Cloud, including Managed Agents (beta), the Files API, the MCP connector and the Console, billed through Google Cloud. It covers Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1, and the Google-managed Vertex option continues.
Why it matters: Google Cloud customers get the same day-one features as Anthropic's own API without a second contract.
Anthropic is donating the Agent Skills specification, name and project to the Agentic AI Foundation. More than 70 platforms, ChatGPT and Cursor among them, have adopted Skills since December 2025.
Why it matters: Skills becomes a vendor-neutral standard, following the same path as MCP.
The new API answers narrow questions such as classify, route or pick a next step, returning a choice, probability or score. It runs on GPT-6 Luna, has no output-token charge and takes up to 400 questions per request.
Why it matters: It makes the many small judgment calls inside agents much cheaper.
CNBC reported that OpenAI is near a round led by UAE investors at a $1.25 trillion pre-money valuation, and Reuters put it above $1.4 trillion. OpenAI has not confirmed the raise.
Why it matters: It would be one of the largest private valuations ever, arriving as Anthropic prepares to go public.
A gradual rollout lets Grok Bot use on-device apps such as Calendar, Reminders, Contacts, Messages and smart-home controls, with permission set per app, on paid plans. The list of apps comes from a secondary report of Musk's post, not official documentation.
Why it matters: Agents that reach personal phone data raise both the usefulness and the privacy stakes.
The program has two tracks: consultancies that help companies deploy Grok, and builders of plugins and integrations. Bain, Databricks, NVIDIA, Snowflake, Booz Allen and Twilio are among the launch partners.
Why it matters: It is xAI's version of the channel programs OpenAI and Anthropic already run.
Claude Duet runs on Opus 5.5 and works on its own: a morning brief, regular checks of your connected apps, and tasks done in a cloud computer that keeps going while your devices are off. It asks before sending messages, paying, signing or deleting, and blocks trading and money transfers outright. US Pro and Max subscribers can join the waitlist, at no extra charge beyond plan usage.
Why it matters: It is Anthropic's answer to ChatGPT Dots and Grok Bot.
Try it: Pro or Max in the US: join the waitlist from the Claude Duet page.
OpenAI's frontier models and Codex will be sold in Google Cloud's Model Garden and Marketplace, counting toward customers' existing Google Cloud commitments. Access is by request now, with general availability promised in the coming weeks. Microsoft remains OpenAI's primary cloud.
Why it matters: OpenAI's models will now be sold on a rival's cloud, next to Gemini and Claude.
New SDK releases, including Python 2.0, TypeScript 1.0 and Go 2.0, default to API version 2026-10-21 and switch from the MIT license to Apache 2.0. Organization webhooks, user profiles and an admin switch to turn off web fetch also arrived.
Why it matters: Developers on the old SDKs need to plan a migration.
Yahoo Finance, citing three sources, reported that Anthropic could file publicly for an IPO within days, with a possible November listing. Anthropic hasn't commented, and the public filing hadn't been confirmed when this page went up.
Why it matters: It could be one of the largest IPOs ever.
Locked chats disappear from history and search and need a four-digit PIN or device authentication to open. They are rolling out on Free, Go, Plus and Pro across web, mobile and desktop.
Why it matters: A simple privacy control for anyone who shares a device.
SpaceX's third-quarter results put its "Super Intelligence" segment, which includes Grok, Grok Bot and Cursor, at about $7.5 billion of $14.6 billion total revenue. Musk said the company is bringing its Colossus 2 data center online.
Why it matters: It is the first clear look at how big xAI's business has become inside SpaceX.
With the X plugin, a Grok Bot can read and reply to X Chat messages, and anyone can tag the official @bot account on X to save or explain posts. On Android it can be set as the default assistant.
Why it matters: Grok Bot now works inside X itself, where xAI has a built-in audience.
Subject to stockholder approval, Anthropic is moving 5% of its shares into a fund for public benefit, which can't be used for lobbying or political giving. The board can add more as the company's value grows.
Why it matters: A major governance move ahead of the company's expected IPO.
Slides makes real-time, shareable presentations that export to PowerPoint, PDF or Google Slides, for Pro, Business and Enterprise users. GPT-6 Pro also got faster and can now run parallel sub-agents.
Why it matters: ChatGPT keeps moving into office software.
Automated pull-request reviews leave research preview, now taking about 6 to 10 minutes instead of about 20. Admins can choose Quick, Standard or Deep reviews per repository through the GitHub App.
Why it matters: AI code review becomes a supported product rather than an experiment.
The new speech-to-speech model is live in the API at $0.08 a minute, and grok-voice-latest now points to it. xAI says it powers Grok Bot's voice mode and ranks first on Artificial Analysis' speech-to-speech index.
Why it matters: Developers get a cheaper, faster voice model for building voice agents.
GPT-4o, GPT-4o mini, GPT-4.1 and GPT-4.1 mini will be shut off in the API, along with fine-tunes of them. OpenAI recommends GPT-6 Luna as the replacement.
Why it matters: Apps still on these models have under three months to migrate.
One plan now covers Grok, the Cursor coding editor and X Premium, in Free, Plus ($8), Premium ($30), Super ($100) and Ultra ($200) tiers. Existing subscribers can keep their plan or switch with a prorated refund.
Why it matters: It resets consumer AI pricing at xAI and folds in Cursor.