The AI Productivity Weekly: Saturday 17 October 2026
Gemini 4 Argon, Meta's Muse Wonder, ChatGPT Slides, a single X plan for Grok and Cursor, and a week of agents moving into the tools teams already use.
This Week in One Minute
Welcome to the first edition of The AI Productivity Weekly, a short Saturday roundup of the AI news that changes how you work.
The theme: agents stopped being separate apps and moved into Microsoft 365, Notion, GitHub and the Gemini app. All dates are in UTC.
1. Google ships Gemini 4 Argon, with scheduled tasks and coding in the Gemini app
Google’s new frontier model is rolling out to Google AI Pro and Ultra subscribers in the Gemini app and AI Mode in Search, and to developers through the Gemini API at an introductory $2/$10 per million input/output tokens until Nov 30, then $4/$20. Argon’s coding agent in Antigravity, inside the Gemini macOS app, is for Google AI Ultra subscribers.
Why it matters for your productivity: if you already pay for Google AI Pro, try handing a recurring workflow to a scheduled task. API builders: the price doubles after Nov 30.
2. Meta launches Muse Wonder for coding and long-running agents
Meta released Muse Wonder on Muse Code and the Meta Model API at $4 input / $20 output per million tokens. The stronger Muse Wonder Full is subscription-only in Muse Code on the Power Usage and new Maximum Usage (80x Everyday) plans. Meta says Wonder Full leads the DeepSWE and SWE-Marathon benchmarks.
Why it matters for your productivity: more competition on price and limits. Test it on your own repo, not the leaderboards.
3. SpaceXSI bundles Grok, Grok Bot, Cursor and X Premium into one plan
“The X Subscription” comes in five tiers: Free, Plus ($8/month), Premium ($30), Super ($100) and Ultra ($200), with one Grok usage pool shared across Grok and Cursor. Premium bundles what were SuperGrok, Cursor Pro and X Premium+, previously $90/month combined. Third-party models in Cursor come only with Premium and above, billed separately at API list prices. Existing subscribers can keep their plan or switch with a prorated refund.
Why it matters for your productivity: if you pay separately for an AI assistant and a coding tool, compare this bundle’s included usage with what you pay now.
4. Notion 3.8 makes your workspace the shared memory for every agent
Any MCP-compatible agent can now read and update your Notion workspace, while admins see and revoke each connection. Notion Agents connect to 150+ tools, a Skills API syncs skills stored in Notion to ChatGPT, Claude and other agents (via GitHub) and databases get column permissions.
Why it matters for your productivity: switch agents without re-explaining context, and share databases with agents without exposing sensitive columns.
5. Microsoft gives Copilot Studio agents their own accounts and memory
Agent accounts let organization-owned agents work in Microsoft 365 with their own credentials. You can @mention them in Teams and Outlook. Business process memory carries context between cycles, and triggers in the GitHub Copilot harness start agents from an email or a schedule. Rollout is through the Frontier program.
Why it matters for your productivity: recurring processes like a weekly budget review can go to an agent that outlasts any one owner, with clear attribution.
6. Anthropic’s Claude Code Review is generally available
Code Review is out of research preview for Claude Team and Enterprise organizations. Anthropic says most reviews now finish about 3x faster, and admins can choose Quick, Standard (the default) or Deep reviews for each repo. Eligible Enterprise orgs can claim Code Review credits by Jan 15, 2027.
Why it matters for your productivity: as AI writes more code, review is the bottleneck. An automatic first pass helps.
7. Perplexity open-sources SPACE, the sandbox behind Computer
SPACE runs every Perplexity Computer task in its own Firecracker microVM and controls credentials and outbound traffic from outside the VM. It’s now available under Apache 2.0 (GitHub). In Perplexity’s earlier evaluation, Escaping SPACE Part I, nine frontier-model setups with root access made 108 escape attempts. None reached the host, though some got around network limits before Perplexity fixed them.
Why it matters for your productivity: teams building agents get a free, tested blueprint for keeping them boxed in, which is what security reviewers ask for.
8. ChatGPT adds Slides, a parallel-agent GPT-6 Pro and new voices
ChatGPT Slides lets eligible Pro, Business and Enterprise users with Space access create decks, edit them, and export to PowerPoint, PDF or Google Slides. GPT-6 Pro now runs multiple agents on different parts of a task in parallel for eligible Pro users. On Oct 15, GPT-Live-1.1 brought more natural accents and new voices to ChatGPT Voice; the voices you see “may vary by region and account.” API access is “Coming soon” (@OpenAI, Oct 15, 18:19 UTC).
Why it matters for your productivity: a conversation can now turn into a shared, editable deck without a copy-paste step into PowerPoint.
Also This Week
- Oct 14: xAI released Grok Voice 3.0, a speech-to-speech model for voice agents, still $0.08 per minute. It tops the Artificial Analysis index, though not on every test.
- Oct 14: Microsoft 365 Copilot Chat now shows memory citations, which you can review or delete (Message Center roadmap).
- Oct 15: ChatGPT’s Chrome extension can now read across your open tabs and, in Work, control your signed-in Chrome, asking before each new site (@ChatGPT). Availability depends on rollout and workspace settings.
- Oct 16: Claude Code 2.1.302 makes fork subagents the default in
claude -p, the Agent SDK, desktop and VS Code. SetCLAUDE_CODE_FORK_SUBAGENT=0to turn them off in scripts or CI. - Oct 16: OpenAI’s API adds prompt-to-voice (a custom voice from a text description) (API changelog).
Tool to Try This Week: Notion + Your Agent via MCP
Pick one agent you already use and connect it to Notion with MCP (setup link in the 3.8 notes). Start with one page: a “How we work” doc with your team’s priorities, style rules and open projects. Ask the agent to read it before each task and to write back what it learns. Admins can revoke it anytime in MCP controls. Check your plan’s feature availability first.
What to Watch Next Week
- Oct 24, 00:00 UTC: Perplexity’s Agent API and Orchestrator API retire seven older OpenAI model IDs: gpt-5.4, gpt-5.4-mini, gpt-5.4-nano, gpt-5.2, gpt-5.1, gpt-5 and gpt-5-mini (changelog). Update any integrations before then.
- The X Subscription switch rolling out to existing SuperGrok, Cursor and X Premium subscribers in billing settings.
- GPT-Live-1.1 in the API, which OpenAI has called “coming soon.”
- Gemini 4 Argon capacity in Antigravity, where Google says it has scaled capacity deliberately.
- ChatGPT dots usage terms: OpenAI’s one-month window from the Sep 29 launch ends around Oct 29, and per-plan terms are still to come.
That’s Edition 1. If one item here saves you an hour next week, forward it to the teammate who’d use it too, and come back next Saturday for Edition 2.