The short version: nine launches and deals from 13 to 26 October 2026 that change what a small business, developer or marketer pays for AI, or what it can hand off. Anthropic opened a waitlist for Claude Duet, an agent app that works while you are away, OpenAI priced its Decisions API at $0.10 per million input tokens, Google gave small business owners a Gemini hub tied to Shopify and Xero, ChatGPT Ads reached 89 countries, and both Claude and OpenAI models can now be paid for from a Google Cloud commitment. Every story links its primary source with the date it was published, and every image is our own screenshot of that page, captured 25 October 2026.
In this edition
- Anthropic’s Claude Duet, a proactive agent app built on Claude Opus 5.5, opens a waitlist preview for US Pro and Max subscribers (21 October 2026)
- OpenAI’s Decisions API reaches general availability at $0.10 per million input tokens, with no charge for output tokens (22 October 2026)
- Google adds a Gemini hub for small business owners that connects to Google Ads, Shopify, Xero and 7 other apps (20 October 2026)
- ChatGPT Ads expands to 26 more countries, reaching 89, after more than 100,000 advertisers ran campaigns (21 October 2026)
- Claude Security scans repositories continuously for Claude Enterprise, with $25,000 in free usage credits a month for two months (21 October 2026)
- Anthropic makes the full Claude Platform generally available on Google Cloud, billed against existing Google Cloud commitments (22 October 2026)
- OpenAI frontier models and Codex come to Google Cloud Model Garden, payable from Google Cloud spend commitments (21 October 2026)
- Anthropic donates the Agent Skills standard, used by more than 70 agent platforms, to the Linux Foundation’s Agentic AI Foundation (22 October 2026)
- Hindsight, an MIT-licensed agent memory system with more than 53,000 GitHub stars, ships version 1.0 and then 1.1 (13 October 2026)
1. Anthropic’s Claude Duet, a proactive agent app built on Claude Opus 5.5, opens a waitlist preview for US Pro and Max subscribers
Anthropic opened a research preview of Claude Duet, an app where Claude messages you first with what needs attention and keeps working on tasks after you step away.
- Claude Duet opened as a research preview by waitlist on 21 October 2026, starting with US Pro and Max subscribers; it is included in those plans and uses the same usage limits as the rest of Claude.
- It runs as a standalone app on iOS and Android and on claude.ai, and Anthropic says it was designed around Claude Opus 5.5.
- It sends a morning brief across your connected apps, checks those accounts every few hours, and has a virtual machine with its own browser, so tasks keep running when your laptop or phone is off.
- Before it sends a message, makes a payment, signs a document or deletes anything, it shows the action and waits for approval; trading securities and moving money between accounts are blocked even with approval.
Vendor claims, not tested by us: A restaurant owner with about 110 employees used it for accounting, customer support and marketing during early access (tester account on Anthropic’s page).
Why it matters: A small business owner already paying for Claude Pro or Max can hand off scheduling, invoice reviews and client call prep at no extra subscription cost, with every send and payment still held for approval.
Source: Introducing Claude Duet, Anthropic, published 21 October 2026.
Our review and score: Claude on AI Review Rating.
2. OpenAI’s Decisions API reaches general availability at $0.10 per million input tokens, with no charge for output tokens
OpenAI made the Decisions API generally available: an endpoint for questions with bounded answers, such as routing a ticket or picking the next step for an agent.
- The Decisions API became generally available on 22 October 2026 through a dedicated Decisions endpoint, built with GPT-6 Luna.
- A developer defines a question and its allowed answers; the API returns a probability (predicate), one of the supplied options (choice) or a rubric level (score).
- Added since the public beta: a native GPT-Live integration, prompt caching, inference residency in the US and EU, native JSON inputs, images in question instructions, and up to 400 questions per request, twice the previous limit.
- Price: $0.10 per million input tokens and $0.01 per million cached input tokens, with no output-token charges.
Vendor claims, not tested by us: Up to 10x faster than GPT-6 Luna through the Responses API (OpenAI). Box reports about 5x faster route selection on 30 certificate of insurance cases, with the same accuracy as the Responses API (customer quote on OpenAI’s page).
Why it matters: A team paying full model rates to classify tickets, route documents or tag transactions can move those yes or no and pick one steps to an endpoint that bills input tokens only.
Source: Build more responsive apps with the Decisions API, OpenAI, published 22 October 2026.
3. Google adds a Gemini hub for small business owners that connects to Google Ads, Shopify, Xero and 7 other apps
Google added a business landing page to Gemini where small business owners manage brand visuals, finances and customer insights, with connections to the apps they already use.
- Google announced a business landing page in Gemini on 20 October 2026 for building a brand, analyzing finances and reading customer insights, rolling out globally from that date.
- Gemini connects to Google Business Profile, Google Ads and Google Workspace, and to Adobe, Canva, Dropbox, Shopify, Wix, Zoho and Xero.
- Owners can upload a website, brand kit, fonts or color palettes; with the Nano Banana 2.1 image model, Gemini keeps that style for later marketing visuals.
- Gemini turns spreadsheets, receipts and purchase records into financial graphics and interactive models.
Vendor claims, not tested by us: Tens of millions of small business owners use Gemini (Google). A ceramics business in Bordeaux saves two to three hours a day on product copy and social posts (Google).
Why it matters: A shop owner on Shopify or Xero can keep brand visuals and bookkeeping questions in one Gemini workspace instead of uploading the brand kit again for every campaign.
Source: New Gemini tools for small business owners, Google, published 20 October 2026.
Our review and score: Gemini on AI Review Rating.
4. ChatGPT Ads expands to 26 more countries, reaching 89, after more than 100,000 advertisers ran campaigns
ChatGPT Ads now run in 89 countries, and advertisers in the new markets can buy them through partners or on their own in Ads Manager.
- On 21 October 2026 ChatGPT Ads expanded to 26 additional countries, for 89 in total.
- The new markets include 14 more Latin American countries, among them Argentina, Chile, Colombia and Peru, plus Nigeria, South Africa, Serbia, Ukraine, Cambodia, Bangladesh and Pakistan.
- Advertisers buy through technology partners or self-service through Ads Manager.
- Ads appear only on the Free and Go plans; Plus, Pro and Enterprise stay ad-free, and OpenAI says conversations are kept private from advertisers.
Vendor claims, not tested by us: More than 1.2 billion people use ChatGPT each week (OpenAI). Over 100,000 advertisers have run campaigns since ads launched just over eight months before the announcement (OpenAI).
Why it matters: A business selling into Latin America, Africa or South Asia can test ads that appear during ChatGPT shopping conversations through self-service Ads Manager, without hiring a partner agency first.
Source: ChatGPT Ads expands to 26 new countries as over 100,000 advertisers have used the platform, OpenAI, published 21 October 2026.
5. Claude Security scans repositories continuously for Claude Enterprise, with $25,000 in free usage credits a month for two months
Claude Security no longer waits for someone to start a scan: it follows your repositories as code changes and tracks each finding until the fix ships.
- The new version of Claude Security entered public beta for Claude Enterprise customers on 21 October 2026; admins enable it in the admin console.
- It runs a deep baseline scan, then scans changes by risk and complexity; repositories connect through the Claude GitHub app, including GitHub Enterprise Server, or an access token from GitLab and other Git providers.
- Findings are marked fixed when the fix lands and reopen if the issue returns; each carries a severity judged against a threat model Claude Security generates from the code.
- It adds an API, webhooks, an MCP connector in Claude.ai and Claude Code, and role-based access controls.
- Organizations receive $25,000 in free usage credits each month for two months, for usage through 31 December 2026.
Vendor claims, not tested by us: Thousands of organizations have found over a million vulnerabilities with it since the public beta began in April (Anthropic).
Why it matters: A Claude Enterprise customer can run continuous code security scanning on up to $50,000 of free usage through 31 December 2026 and compare it with its current scanner before paying.
Source: Claude Security now scans continuously and tracks fixes, Anthropic, published 21 October 2026.
Our review and score: Claude on AI Review Rating.
6. Anthropic makes the full Claude Platform generally available on Google Cloud, billed against existing Google Cloud commitments
Google Cloud customers can now use the full Claude Platform with their Google Cloud logins and invoice, run by Anthropic rather than Google.
- Claude Models on Google Cloud (Anthropic managed) became generally available on 22 October 2026, with US and Global endpoints.
- It uses existing Google Cloud IAM credentials, bills on a single Google Cloud invoice, and draws down an existing Google Cloud commitment.
- It carries the full Claude API, including Managed Agents (beta), code execution, computer use, agent skills, the Files API, web search, prompt caching and batch processing, plus the Claude Console, with new features shipping the same day as on the native platform.
- Fast Mode on Claude Opus 5.5 is generally available there at $8 per million input tokens and $40 per million output tokens.
- The Google managed offering stays available for customers who need Google Cloud as the data processor or strict regional data residency.
Vendor claims, not tested by us: Fast mode generates output tokens up to 2.5x faster (Anthropic).
Why it matters: A company with unspent Google Cloud commitment can pay for Claude agents out of that budget; Anthropic says to contact your account executive first, because existing discounts cannot be applied retroactively.
Source: The full Claude Platform, available on Google Cloud, Anthropic, published 22 October 2026.
7. OpenAI frontier models and Codex come to Google Cloud Model Garden, payable from Google Cloud spend commitments
OpenAI and Google Cloud are bringing OpenAI frontier models and Codex to Google Cloud’s Model Garden, bought through the Google Cloud Marketplace.
- OpenAI and Google Cloud announced on 21 October 2026 that OpenAI frontier models and Codex agentic workflows are coming to Google Cloud’s Model Garden.
- Customers will buy them through Model Garden, transacted through the Google Cloud Marketplace, and can draw down existing Google Cloud spend commitments.
- OpenAI says its models will also be available in Google Antigravity and the Gemini Agent.
- No general availability date was given on 21 October 2026; an access request form is open, and Microsoft remains OpenAI’s primary cloud partner.
Why it matters: A business that runs on Google Cloud will be able to buy OpenAI and Claude models on the same invoice and commitment, so choosing a model no longer means signing a second cloud contract.
Source: OpenAI models are coming to Google Cloud, OpenAI, published 21 October 2026.
8. Anthropic donates the Agent Skills standard, used by more than 70 agent platforms, to the Linux Foundation’s Agentic AI Foundation
Anthropic handed Agent Skills, the SKILL.md folder format agents load for specific tasks, to the same Linux Foundation group that already runs MCP.
- On 22 October 2026 Anthropic donated Agent Skills to the Agentic AI Foundation under the Linux Foundation, which it co-founded with Block and OpenAI and which already holds the Model Context Protocol.
- The donation covers the Agent Skills name and logo, the GitHub organization and the agentskills.io domain; anyone can use or contribute to the standard without joining the foundation.
- Engineers from Anthropic, Google, OpenAI and other companies maintain it as individuals, with David Soria Parra and Mahesh Murag as lead maintainers.
- The proposed roadmap covers versioned releases, a formal proposal process and security guidance for people who write and install skills.
Vendor claims, not tested by us: More than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot and VS Code, have integrated Agent Skills; 1 million skills have been published and nearly 280 million installed (Anthropic).
Why it matters: A team that writes skills for its own workflows can reuse the same SKILL.md folders across Claude, ChatGPT, Cursor and Copilot under a vendor-neutral standard instead of rebuilding them for each tool.
Source: Donating Agent Skills to the Agentic AI Foundation, Anthropic, published 22 October 2026.
9. Hindsight, an MIT-licensed agent memory system with more than 53,000 GitHub stars, ships version 1.0 and then 1.1
Hindsight is an open-source memory layer that lets AI agents keep and reuse what they learn across sessions, and it reached version 1.0 on GitHub.
- v1.0.0 was published on 13 October 2026 and v1.1.0 on 21 October 2026 (GitHub API, read 25 October 2026).
- 53,530 stars and 6,240 forks on 25 October 2026; forks per star 0.117, above the 0.02 line where bought stars tend to sit.
- MIT licence; the README covers retain, recall and reflect operations, an MCP server and coding agent integrations, with a hosted Hindsight Cloud as an alternative to self-hosting.
- v1.1.0 adds the OpenAI Decisions API and Cloudflare Clef as decision-model providers.
Vendor claims, not tested by us: State-of-the-art results on the LongMemEval benchmark as of January 2026 (Vectorize, in the README).
Why it matters: A team building support or sales agents can give them long-term memory with self-hosted MIT-licensed code before it pays a hosted memory service per call.
Source: vectorize-io/hindsight release v1.0.0 on GitHub, Vectorize (Hindsight), released 13 October 2026. Also: Hindsight v1.1.0 release notes on GitHub.
Our review and score: Hindsight on AI Review Rating.
How we picked and checked these stories
We lead with launches and tools a business can use or price, and leave out stories that only describe risk unless they change what a buyer should do. Creator posts and videos are where we hear about a story; they are never the source. Every fact above comes from the company’s own page, the GitHub API or a primary news outlet, read on 25 October 2026, and every story is dated inside 13 to 26 October 2026. Numbers we could not test ourselves are labelled as vendor claims. Stories we could not verify on a primary source in that window were cut. More editions: all AI News editions.
Your turn
Which would you hand work to first: an agent app that messages you when something needs you, or a cheap API that makes the small routing calls inside your own software? Reply to our newsletter and tell us; the answers shape what we test next.
Get the next edition by email: join our free newsletter with the new AI tools we score.
Related reading
- AI News, week of 19 October 2026: the previous edition
- Our weekly AI tools digest for 19 October 2026: the tools we scored
- Open-source alternatives to paid AI tools
Sources
- Introducing Claude Duet (Claude, 21 October 2026), read 25 October 2026.
- Build more responsive apps with the Decisions API (OpenAI, 22 October 2026), read 25 October 2026.
- New Gemini tools for small business owners (Google, 20 October 2026), read 25 October 2026.
- ChatGPT Ads expands to 26 new countries as over 100,000 advertisers have used the platform (OpenAI, 21 October 2026), read 25 October 2026.
- Claude Security now scans continuously and tracks fixes (Claude, 21 October 2026), read 25 October 2026.
- The full Claude Platform, available on Google Cloud (Claude, 22 October 2026), read 25 October 2026.
- OpenAI models are coming to Google Cloud (OpenAI, 21 October 2026), read 25 October 2026.
- Donating Agent Skills to the Agentic AI Foundation (Anthropic, 22 October 2026), read 25 October 2026.
- vectorize-io/hindsight release v1.0.0 on GitHub (published 13 October 2026), read 25 October 2026.
- Hindsight v1.1.0 release notes on GitHub, read 25 October 2026.
- GitHub REST API, repository and release records for vectorize-io/hindsight, read 25 October 2026.
Comments
Sign in to join the discussion.
Login to comment