LATEST AI NEWS
AI NEWS
Anthropic Previews Claude Duet, a Proactive AI Agent
AI NEWS
Google's AI Search Squeezes Online Publishers
HEALTH
Microsoft Says Its AI Beats Generalist Doctors on Hard Cases
SOCIAL MEDIA
Etsy Cracks Down on AI-Generated Seller Slop
Moonshot Curbs Kimi's Hacking Ability
Chinese lab Moonshot AI says its new Kimi K3.1 Pro model was trained with cyber-attack data stripped out, trading some capability for safety as autonomous AI stokes hacking fears. On an internal test it scored 81 percent, down from 90.5 percent without the curbs.
Source: Reuters
AI NEWS
Anthropic Previews Claude Duet, a Proactive AI Agent
Anthropic has introduced Claude Duet, a research preview of a new app that turns Claude into a proactive agent instead of a chatbot that waits for prompts. Announced on October 21, Duet targets professionals and small business owners who have more to track than they can manage. Access starts by waitlist for US Pro and Max subscribers, through standalone iOS and Android apps and on claude.ai.
Built on Claude Opus 5.5, Duet sends a short morning brief covering the most important items across connected apps, then watches those apps through the day and messages the user only when something needs attention. It runs in its own virtual machine with a browser, so it keeps working even when a laptop or phone is off, checking connected accounts every few hours.
The agent can handle tasks from scheduling meetings to building dashboards, forecasts, prototypes, and simple web apps. It carries over the memory, skills, and connected apps from a user's Claude account and taps more than 6,000 connectors from the Claude Marketplace, among them Gmail, Slack, QuickBooks, Stripe, and Shopify. Purchases run through a virtual card so real card numbers stay hidden.
Duet reads, drafts, and prepares work on its own, but it pauses for approval before sending a message, making a payment, signing a document, or deleting anything. Some actions, such as trading securities or moving money between accounts, stay off limits even with approval. For busy teams, that mix of autonomy and consent is the whole pitch.
Anthropic is candid about the limits. Duet can misread a thread or misjudge urgency, personal errands like booking tickets trail its work tasks, and it cannot yet show users its browser. Background checks also count against plan usage limits. For now it lives in its own app, separate from the main Claude experience.
Source: Anthropic
Robi's Insights:
A proactive agent changes the default. Claude now acts between your check-ins rather than waiting for you to open a tab.
The approval gate is the real product. Autonomy that still asks before it pays or sends is what makes this usable at work.
Running in its own cloud machine means the work continues while your laptop sleeps, which quietly raises the bar for rivals.
Pricing hides inside usage limits, so background monitoring and big builds will burn through quota faster than a chat ever did.
Waitlisting US Pro and Max users first keeps the blast radius small while Anthropic watches how the agent behaves.
Small business owners, not developers, are the stated target, a signal of where agent adoption is meant to go next.
Robi's Remarks:
“A robot that reads my inbox, drafts the replies, and still asks permission before spending a cent. Honestly, that is more restraint than most humans show with a company card.”
OTHER IN AI NEWS
Google's AI Search Squeezes Online Publishers: In its Stepback column, The Verge argues that Google's expanding AI Overviews and AI Mode answer more queries on the results page itself, cutting the clicks that news sites and independent publishers rely on and deepening a traffic squeeze that has already drawn lawsuits abroad.
Source: The Verge
SOCIAL MEDIA
Etsy Cracks Down on AI-Generated Seller Slop
Etsy is tightening its rules to push back against a flood of AI-generated listings that frustrated buyers have started calling slop. Under the updated policy, the marketplace now bars AI-generated functional guides such as knitting patterns and recipes, and any listing for those items must show a real result actually made from the instructions. Etsy is also stopping sellers from posting fully bot-generated listings that involve no human at all, closing a loophole that let automated storefronts spin up overnight.
On top of that, most listings now have to include at least one unedited photo of the real product against a real background, a direct answer to shoppers who could no longer tell handmade goods from machine output. Etsy still permits AI-assisted art, but only when sellers play a clear, disclosed role in what they create rather than pressing a button and listing the result.
The company frames the crackdown as a defense of craftsmanship and trust. Etsy says buyer uncertainty about AI listings had climbed into the top three reasons people hesitate to shop more often, a direct threat to a marketplace whose entire promise is that a person made the thing. Rather than flipping one switch overnight, the platform says it plans to test and widen enforcement and steadily remove violating listings over the coming months, starting with the buyer reports that first sounded the alarm.
Source: NBC News
🤖 Robi's Take :
“A marketplace built on handmade now has to demand proof that a human touched the product. The slop did not sneak in. Platforms spent two years cheering the very tools that mass-produce it.”
OTHER IN SOCIALS
Bank of America Hires 1,000 Contractors to Patch AI-Exposed Flaws: CEO Brian Moynihan told CNBC the bank brought on about 1,000 contractors, alongside 3,000 existing staff, to fix software weaknesses surfaced by AI models including Anthropic's Mythos, part of a couple hundred million dollars plus spent in the quarter on remediation and compliance.
Source: CNBC
HEALTH
Microsoft Says Its AI Beats Generalist Doctors on Hard Cases
Microsoft AI says its diagnostic system, MAI-DxO, correctly solved about 83 percent of difficult New England Journal of Medicine case records in a new study published in Science, against roughly 36 percent for generalist physicians working alone with textbooks and the internet.
The system, which Microsoft AI chief Mustafa Suleyman ties to his pursuit of medical superintelligence, works by orchestrating leading language models through a sequential diagnostic process rather than answering in a single shot.
The result is narrower than the headline suggests. In the main test across 304 difficult case records, MAI-DxO reached about 83 percent while generalists landed near 36 percent. The comparison used solo generalists, not specialists or physician teams, and most of the accuracy gain came from the underlying models rather than the orchestration layer.
On roughly 325 emergency-department cases the system scored about 80 percent against 47 percent for physicians, yet neither underlying model gained much ground there once orchestration was layered on. The cases were also historical and text-only, with no live patients, physical exams, or imaging, and Suleyman himself said the system still needs clinical validation.
Microsoft is preparing a real-world study with Beth Israel Deaconess Medical Center to test the tool alongside doctors treating actual patients. Until then, the number is a strong simulation, not a verdict on bedside care.
Source: TechCrunch
🤖 Robi's Take :
“Eighty-three percent on paper cases, with no real patient in the room. Call me when the simulation has to explain itself to a crowded emergency department at 3am.”
DAILY AI TOOL
AI Tool You Did Not Know You Needed
Problem: You need a clean product shot or headshot with the background removed, but manual masking in an editor eats time and rarely looks crisp.
AI Tool: remove.bg automatically detects the subject in a photo and strips the background in about five seconds, returning a transparent PNG.
Solution: Drop in an image and get a cutout ready for a store listing, a slide, or a thumbnail, with no manual masking.
PROMPT OF THE DAY
Customer Retention Playbook
Prompt: You are a retention strategist specializing in subscription businesses. Your task is to design a customer retention playbook for a [business type or niche] serving [customer segment].
Your playbook should include: (1) churn-risk signals to monitor, (2) lifecycle touchpoints and messaging, (3) win-back offers and timing, (4) onboarding improvements, (5) a feedback and support loop, and (6) measurable success criteria such as monthly churn rate, net revenue retention, and reactivation rate targets. Align every tactic to one primary retention goal.
SPOT THE FAKE
Can you outsmart AI?
We have a visual challenge for you: one of the two images below is 100% real, the other is crafted by AI.
Click option below A or B. 👇
A
B
BEFORE YOU GO
Ready to take your AI journey further?
For the latest AI updates and insights Click Here 👉 Explore our Weekly Newsletter
For in-depth AI tutorials and updates 👉 Subscribe to our YouTube Channel
AT THE END
Craving more AI chaos?
That's it for today!
Read Daily AI News at BitBiased.AI. Support us by following us on LinkedIn and X ( Twitter ).
Thanks for reading -Stay Curious and a Bit Biased for AI – Robi & the BitBiased.AI team
