Agents Everywhere: Anthropic’s Proactive AI, OpenAI on Google Cloud, and Walmart’s Bot Shoppers

From an assistant that works before you ask, to a cloud truce that would have been unthinkable two years ago — today’s tech news has one theme: AI agents are moving from answering questions to doing the work.

Anthropic’s Claude Duet: An Assistant That Works Before You Ask

Anthropic unveiled Claude Duet, a standalone AI agent app for iPhone, Android, and the web at claude.ai. The key word is proactive: where most AI waits for you to type something, Duet monitors your connected apps and surfaces what needs your attention on its own. Every morning, it hands you a briefing of what’s coming at you — and then it does the work.

That work includes scheduling meetings, building financial forecasts, and even building web apps. Anthropic says Duet integrates with more than 6,000 connectors in the Claude Marketplace — Gmail, Slack, and the usual workplace tools — and it can even make purchases using Stripe’s Link wallet and Shopify’s Shop Pay.

With that much power, Anthropic says Duet requires user approval for sensitive actions. To its credit, the company acknowledged that the agent can misjudge urgency and remains less effective at personal tasks like booking tickets — honest framing for an agent product. Duet is rolling out as a waitlisted research preview to U.S. subscribers on the Pro and Max plans, built around Anthropic’s Opus 5.5 model, sharing existing usage limits.

The competitive timing is hard to miss: OpenAI launched its always-on agent last month, and Meta’s Muse agent pulled millions of downloads. Everyone is racing to be the assistant that acts before you’re asked.

OpenAI’s Models Move Into Google Cloud

In a deal that would have sounded absurd two years ago, OpenAI’s frontier models are coming to Google Cloud. The cloud provider announced that models including GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna will be available in its Model Garden. Enterprises buy them through the Google Cloud Marketplace — and here’s the big incentive: they can pay using their existing cloud spending commitments.

That removes real procurement friction: a company already committed to spending millions a year on Google Cloud can now spend some of that commitment on OpenAI models instead of negotiating a separate contract. Regulated industries get OpenAI models wrapped in Google’s access controls and data governance, which matters for building agentic workflows in banks, hospitals, and similar environments.

The irony is delicious. Google and OpenAI compete in every AI product that matters, and now Google is reselling its rival’s models. OpenAI’s exclusive cloud arrangement with Microsoft expired in April, and since then the company has expanded to Amazon Web Services and Oracle — and now Google. OpenAI models now run on all three big American clouds, even as the company says Microsoft remains its primary cloud partner. Functionally, the cloud wars have become model wars, and the clouds are all deciding to be neutral shelves.

Walmart Rolls Out the Welcome Mat for AI Shoppers

Walmart announced tools that let AI agents shop and complete purchases for customers, rolling out something called Agent Profiles next month across Walmart and Sam’s Club accounts.

The idea is simple: you authorize a specific agent — ChatGPT, Gemini, or Meta’s Muse — set a spending limit or require your approval for each purchase, and you can revoke access at any time. The agent’s orders stay separate from your own, so you can see exactly what the machine bought. No Walmart+ membership required, just a regular account.

Under the hood, Walmart built the Walmart Agent Gateway on the Model Context Protocol, the open standard Anthropic created. Instead of an agent clicking through web pages built for humans, it talks to Walmart’s systems directly to search inventory and check out.

The guardrails are the interesting part:

• Agents never see your payment card information, and they can’t add or change payment methods.
• If an agent orders the wrong thing, Walmart’s normal returns still cover you.
• Agents that break the rules get blocked outright.

Contrast this with Amazon, which has been blocking third-party AI shopping agents from its platform. Two philosophies: Amazon says keep the bots out; Walmart says come through my door, on my terms. Walmart is betting that trust, not walls, wins the agentic shopping era.

ChatGPT Ads Go Global: 89 Countries and Counting

How is OpenAI paying for all of this? Ads. ChatGPT advertising is expanding to 26 more countries, bringing the total to 89. More than 100,000 advertisers have run campaigns since launch earlier this year, and the business hit an annualized revenue run rate of $1 billion in August — less than 200 days after launch.

The expansion leans heavily on Latin America, with 14 more countries including Argentina, Chile, Colombia, and Peru, plus markets across Europe, the Middle East, Africa, and Asia, such as Nigeria, South Africa, Serbia, Ukraine, Cambodia, Bangladesh, and Pakistan.

The line OpenAI is holding: ads only appear on the Free and Go plans. Plus, Pro, and Enterprise subscribers stay ad-free. It’s the classic freemium play — and with a potential initial public offering ahead, that advertising revenue line is becoming a bigger part of the story.

Legal AI’s Mega-Round: Legora Hits an $8.8B Valuation

Meanwhile, in the venture world, legal AI is having its moment. Legora, the Stockholm-based startup, just raised $550 million at an $8.8 billion valuation, led by Atreides Management. This is a Series E only about seven months after the company’s last round, which valued it at $5.6 billion.

Revenue is the story: annual recurring revenue is over $200 million, after Legora went from $1 million to $100 million in under 18 months. Its biggest competitor, Harvey, raised the same amount — $550 million — just last month, at a $15.5 billion valuation. Legal work, with its drafting, research, and review, is one of the clearest fits for AI agents, and investors are pouring in.

The Chips Underneath It All

A quick hit to close: Google Cloud’s new A5X instances, built on Nvidia’s Vera Rubin racks, go generally available in November, announced at GTC Berlin. Google says it plans to scale that infrastructure to as many as 960,000 Rubin GPUs across multiple sites, aimed at agentic inference and reinforcement learning. The first customer is Ineffable Intelligence, the London lab founded by AlphaGo creator David Silver. Every agent in today’s episode needs somewhere to run — and it’s going to be Rubin racks.

Sources

• Anthropic launches Claude Duet proactive agent app: Reuters
• OpenAI’s frontier models coming to Google Cloud: Reuters
• Walmart opens to AI shopping agents with Agent Profiles and Agent Gateway: Reuters
• ChatGPT ads expand to 26 more countries, 89 total: Reuters
• Legora raises $550M Series E at $8.8B valuation: Reuters

Adapted from the October 22, 2026 episode of Daily Tech and AI.

Daily Tech and AI is AI-voiced and produced by Angelo Ruiz.