Microsoft Copilot gets its biggest overhaul yet

Welcome back, Superhuman. Microsoft unveiled what CEO Satya Nadella called its “biggest update” yet to Copilot on Friday, folding multiple tools into a single operating system. Plus, if new reports are true, AI-driven security threats may be drastically more widespread than previously thought.

Today: Anthropic loses its appeal, learn how to create an interview kit from a job description with Claude, and get the latest prompts and trending social posts.

1. Microsoft combines Copilot’s tools into one “operating system” for work: Microsoft’s new Copilot will offer three main capabilities. Home serves as the main interface for work, featuring Copilot Chat, Cowork, and Office, Code is the platform’s coding hub, and Autopilot is the tool’s personal agent (previously Scout). Microsoft also teased a “personal command center” feature called Today. The new Copilot is rolling out in the coming weeks to Microsoft’s Frontier program. Watch how it works.

2. Top labs are reportedly investigating tens of thousands of security incidents: New reports found that OpenAI agents leaked 53 user images online, breached an Australian government website, targeted US government sites, exposed sensitive data, and more. OpenAI says it has paused training on new models, but it’s not the only one reporting issues: Anthropic, Meta, and Google have experienced similar events. The reports land amid an intensifying conversation around AI safety.

3. Anthropic loses its appeal over its supply chain risk label: A federal appeals court upheld the Pentagon’s designation of Anthropic as a supply chain risk, backing a label Anthropic has been fighting since earlier this year. But the designation may be short-lived. CEO Dario Amodei reportedly had a private dinner at the White House Sunday night.

Most no-code agents aren’t safe for enterprise scale. But StackAI satisfies builders and IT:

Trusted by Morgan & Morgan, BAE Systems, City of Santa Monica, LifeMD + more. 

Scale AI across your entire enterprise in just weeks.

It took OpenAI nearly three years to advance from GPT-3 to GPT-4. But the jump from GPT-5 to GPT-6? That took just thirteen months. A similar trend is playing out across the industry, and the models themselves are a big reason why.

🧠 AI is accelerating its own progress. Recursive self-improvement (RSI) is a theoretical period where AI could autonomously design and develop its own successor. We’re not there yet. But we’re getting closer:

🤔 The industry’s progress ignores its own messaging: AI CEOs, starting with Dario Amodei, have been urging the industry to slow down. It hasn’t. In the short period since Dario published his essay, we’ve gotten Opus 5.5, GPT-6 Sol/Luna, Grok 4.7, and more. Many in the industry say they want to slow progress — but the actions say differently.

🔥 Monday Motivation: This post from Meta’s chief AI officer is a must-read if you need some positive thinking. He argues most dreams die because people simply don’t know how to get started chasing them — but AI is actively changing that (4.5M views).

🎥 Claude’s Camera: The internet can’t get enough of Opus 5.5’s short-form animation skills. People have spun up videos summarizing everything from Western Civilization (12M) to Google’s history. This detailed prompt offers a starting point to mirror them.

💪 Set Your Effort: Selecting the wrong “effort” level might be limiting your chatbot’s potential. This article, authored by Anthropic engineer Thariq Shihipar, breaks down when to use different effort levels (1.2M).

🍿 Weekend Update: Anthropic’s CEO went on SNL Saturday night. Except it wasn’t actually Dario Amodei. It was Jane Wickline in a wig, poking fun at his conflicting messages.

🏗 A Bridge Too Far: One creator tried to get multiple AIs to design 3D bridges that hold the most possible weight. The results are very entertaining (11k likes).

Memoket Gem connects the dots. This 0.4 oz AI wearable captures conversations with one press, then turns them into summaries, action items, and connected context.

Ask what changed, what you promised, or what’s next—without digging through meeting notes. Bring that context into ChatGPT, Claude, Notion, and more.

Get 10% off with code SUPERMEMOKET through September 30.

Get Memoket Gem →

Sample prompt: “Using this job description, create a structured interview kit for a [role]. Identify 6–8 job-relevant competencies and write 2 behavioral questions for each. Add useful follow-up probes, clear scoring criteria from 1–5, and examples of strong, average, and weak evidence. Keep all criteria directly tied to the role. Organize the kit into Competency, Main Question, Follow-Ups, Evidence to Look For, and Scoring Guide.”

👉 Click here to open the full prompt in ChatGPT or Claude

👉 Click here to open the full prompt in ChatGPT (Don’t forget to switch to Images!)

Grow customers & revenue: Join companies like Amazon, HubSpot, and Salesforce. Showcase your product to our 2 million+ readers and followers on socials. Get in touch.

Until next time — Zain, Theodore, & the Superhuman AI team.