It's Wednesday, October 28, and the news is good for you.
The Athena Brief is a weekly AI newsletter keeping you up-to-date with the world's fastest changing (and most impactful) industry, as well as tips and tricks to save you real time & make you real money using AI.
This week starts with your to-do list
Anthropic's New Claude Duet Doesn't Wait for You to Ask
This assistant checks in before you do
Anthropic opened a research preview of Claude Duet on October 21. It's a new app where Claude checks your connected accounts every few hours, completes the tasks you hand it, and keeps working after you close the app.
Each morning it sends you a short brief on the most important items across your apps. Before it sends a message, makes a payment, signs a document, or deletes anything, it shows you exactly what it's about to do and waits for your yes.
Anthropic built it for professionals and small business owners "with more on their plates than they can track." For now it's a waitlist for US Pro and Max subscribers, included in those plans, on iOS, Android, and claude.ai. The work it does in the background counts toward your plan's usage limits.
It's been a busy fall for AI agents. xAI launched Grok Bot on August 11 as "AI teammates" for work. Meta followed with Muse, a personal AI agent, on September 8, and OpenAI with dots, an assistant inside ChatGPT for Pro and Business Premium users, on September 29. Grok Bot is pitched at work teams, Muse at your personal life, and dots at people who message it from work apps like Slack and Teams. Duet is the one pitched at small business owners.
Why it matters
The regular Claude app waits for you to open a chat. Duet shows up on its own. Every few hours it sweeps your accounts like a searchlight, and the invoice that slipped past you gets caught.
It plugs into most apps small businesses already run on: QuickBooks, Stripe, Gusto, HubSpot, Gmail, Slack, Shopify, and Docusign. The examples Anthropic gives are the jobs you'd hand off first: check QuickBooks or Stripe for unpaid invoices and draft reminders, draft follow-ups from your HubSpot deal history, and find the Gmail and Slack messages you missed.
What to do
- List what didn't need you Write down the tasks you did twice last week that didn't need you. The Tip of the week below turns that list into a job description any assistant can work from.
- Set your approval lines Decide now what an assistant may never send, pay, sign, or delete without you.
Between the lines
Duet never sees your passwords, and nothing that sends, pays, signs, or deletes goes out without your approval.
That's the bar to hold any AI assistant to once it's inside your invoices. Anthropic says plainly that Duet still might misread a thread or miscalibrate what's urgent, so keep your approval on everything that leaves the building.
Dive deeper
- Introducing Claude Duet, from Anthropic
- Claude Duet Safety, Privacy, and Security, from Anthropic
- Claude Duet: Anthropic previews an AI agent that messages you first, from VKTR
Quick hits
Google is putting ads in AI Mode's follow-up questions
Google said on October 28 that it's launching multiturn ads in the U.S., so ads can show as a customer keeps asking AI Mode questions, not just on the first search. They're labeled "Sponsored."
If you already run Search, Shopping, or Performance Max campaigns, you're automatically eligible. You can't opt out of AI Mode separately from Search, but your negative keywords and brand exclusions still apply there.
Sound familiar? OpenAI started testing ads in ChatGPT in February, after a backlash over app suggestions that looked like ads. And remember Anthropic's banger Super Bowl ad earlier this year? Its point: Claude will stay ad-free.
Anthropic filed to go public
Anthropic filed its S-1 on October 26 to list on the Nasdaq as "ANT." Fortune puts the expected valuation at $1.8 trillion to $2.0 trillion.
- Revenue: $16.3 billion in the first half of 2026, up from $4.6 billion in all of 2025.
- Operating loss: $693 million, on $17.0 billion in operating costs.
- Net loss: $80.7 billion for the half. $72.4 billion of that is a non-cash paper charge, mostly from convertible notes held by investors that went up in value as Anthropic's valuation rose.
The share count and price range haven't been set.
OpenAI's GPT-6.1 Terra does more for half the price
The same day Anthropic filed, OpenAI cut prices. GPT-6.1 Terra came out October 26 in ChatGPT Work, Codex, and the API, priced 50% below the Terra it replaces. It's built for everyday work, from drafting proposals and reading technical drawings to filling out forms.
On OpenAI's internal test of prompts built to trip it up, its error rate at low effort fell to under a third of the old Terra's.
Tip of the week
Write your AI assistant a job description
Give your assistant a brief, not a hunch.
- List every task you did twice last week that didn't need you, like invoice reminders, follow-ups, and rescheduling.
- Next to each, write where the information lives: your accounting app, your inbox, or your CRM.
- Set your permission level for each task: always act alone, ask for my permission, or never do.
- For each task, write one line on what "done" looks like.
- Paste the list into the assistant you use and ask it to draft the first task for you to check.
Run the first week with you checking every answer. Hand off the tasks it gets right and refine the ones it doesn't.
Or, if this sounds like a drag, the team at Athena would be happy to help.
One number
people in the U.S. and Canada who had set up Muse, Meta's personal AI agent, by October 21, the day Duet opened its waitlist
|
|
P.S. from Athena: Wrote the job description and not sure which task to hand off first? Chat with me and we'll pick it together. |