Hey — Here’s last week AI roundup.

It’s been a year since we last published this newsletter, and we’re excited to finally be back.
And this time we’re committed to staying consistent. Our promise is simple: to deliver clear, useful, and valuable AI updates regularly, without unnecessary noise or long gaps. thank you.

In today’s AI researches:

• Anthropic’s Mythos Broke Into Almost Every NSA Classified System in Hours

• Meta employees are reportedly frustrated after being reassigned to train AI models instead of building products.

• Midjourney unveiled a 60-second, radiation-free full-body ultrasound scanner

• Learn how to Conduct better stock research with Perplexity Finance

• AI Researches case studies

• AI TIP of this week

• wild predictions from Anthropic CEO, Satya Nadella & Google’s former CEO

• Best performing Ai tools of the week

• More Ai news and developments of this weekends

Read time: 6 minutes

 TRENDING 

🔓 Anthropic’s Mythos Broke Into Almost Every NSA Classified System in Hours

AI Researches: The NSA confirms Mythos “broke into almost all of our classified systems, not in weeks, but in hours,” according to testimony shared during a Senate Intelligence Committee hearing.

The Details:

• Senator Mark Warner said NSA and Cyber Command chief General Joshua Rudd told him Mythos broke into almost all of their classified systems within hours.

• Soon after, the U.S. government ordered Anthropic to limit Mythos 5 and Fable 5 to American users, which forced both models offline worldwide.

• Anthropic only shared Mythos with around 200 trusted partners, and says it found thousands of software flaws, including one hidden for 27 years.

• The shutdown also blocked close U.S. allies and overseas safety teams, showing that powerful AI models are now becoming national security tools.

 iMajor AI Developments that Happened in this week i

Meta employees are reportedly frustrated after being reassigned to train AI models instead of building products.
The shakeup follows 8,000 layoffs, affects thousands of workers, and comes as the company's AI costs are expected to reach billions of dollars.

A Chinese AI model just caught up to some of the world's best closed-source systems.
Z.ai's new GLM-5.2 is topping coding and agent benchmarks, narrowing the gap between open models and industry leaders like GPT and Claude.

Anthropic launched Artifacts for Claude Code, turning AI coding sessions into shareable web pages.
The feature lets teams share live dashboards, code changes, charts, and project updates through a private link.

Sakana AI launched Fugu, a multi-agent system designed to avoid dependence on restricted U.S. models.
It routes tasks across multiple open and closed models through one API, with Fugu Ultra claiming performance close to top restricted models.

Genesis AI just unveiled Eno, a headless robot built for real-world work.
The general-purpose robot features dual arms, adjustable height, and AI-powered reasoning designed for tasks ranging from warehouse operations to household chores.

 TRENDING 

⛵ Midjourney unveiled a 60-second, radiation-free full-body ultrasound scanner

AI Researches: Midjourney just unveiled Midjourney Medical, a new water-based ultrasound scanner that aims to create detailed 3D body maps in under a minute, with the long-term goal of making scans faster, cheaper, and easier to access.

The Details:

• The scanner uses a shallow pool of water and a ring of around half a million tiny ultrasound sensors to send sound waves through the body from every angle.

• Midjourney says the system can produce MRI-like 3D images at much higher speed, with the goal of scanning a full body in about 60 seconds instead of a long hospital-style appointment.

• The prototype is being built by a small nine-person team and reportedly processes massive scan data in real time, around 17GB per second, which is pretty wild for a medical device.

• The company is not just imagining this as hospital equipment, but more like a relaxing spa-style scan location, with plans pointing toward public sites as early as 2027.

💰 Conduct better stock research with Perplexity Finance

Use Perplexity Finance to screen stocks, review financial data and ask source-backed questions from one workspace

1. Open Perplexity Finance
   Download Comet, open a new tab and select Finance from the top menu.

   2. Find stock ideas
   Open Screener and enter a broad request, such as: “Find energy stocks with low P/E ratios.”

   3. Research each stock
   Select a ticker and review Overview, Financials, Earnings and Analysis for key information.

   4. Ask the AI assistant
   Use the Comet assistant at the top right to investigate price movements and supporting evidence.

   5. Verify and track
   Open the cited sources, confirm their dates and add promising stocks to your watchlist.

   Prompt:
   “Explain today’s stock move in plain English. What are the likely drivers, and what sources support each one?”

A new Pew Research survey found that 49% of Americans now use AI chatbots, up from 33% in 2024, but skepticism remains strong. Nearly two-thirds believe AI is advancing too quickly, and while many use it for work and information search, most remain concerned about its impact on society, privacy, and regulation.

According to Forbes’ 2026 Billionaires List, 35 billionaires are under 30 years old, but only 12 are self-made, with most building their fortunes in AI, tech, and prediction markets. Notable names include the founders of Mercor, Cursor, Lovable, and Kalshi, highlighting how AI startups are creating a new generation of young billionaires.

A new PwC study found that companies using AI to enhance employees rather than replace them are growing faster, hiring more, and paying higher wages. AI-focused firms saw 52% headcount growth vs. 36% for less AI-exposed companies, while jobs requiring AI skills are growing 8x faster than the overall job market, with increasing demand for human skills like judgment, creativity, and leadership.

⏰ Automate Updates with ChatGPT Scheduled Tasks

Create a task
Open ChatGPT → Go to the Scheduled page in the sidebar → Create a new task and choose when it should run.

Use natural language
You can also create tasks by chatting with ChatGPT. For example: “Before every Portugal match, send me a scouting report on the opponent.”

Automate recurring updates
Schedule one-time or recurring tasks, or ask ChatGPT to monitor changes and notify you when there’s a meaningful update. Manage everything from the Scheduled page.

🔗 Learn more→

Google’s former CEO says AI is making it easier to build wealth, but only for people who understand tools like agents, Claude Code, prompts, memory, MCP, and workflows.

Satya Nadella says every company will need both human capital and AI capital.
The Microsoft CEO argues that businesses should focus on turning their expertise, workflows, and institutional knowledge into AI systems that continuously improve over time.

Anthropic CEO Dario Amodei warns AI is advancing faster than governments can respond.
In a new essay, He argues that governments need stronger oversight now as AI systems gain capabilities that could impact cybersecurity, biology, and global power.

• 🩸 Lucis — Tracks 100+ biomarkers from blood, urine, and saliva twice a year, then gives you a personalized action plan from real clinicians.

• 🏠 Drafted — Generates custom house plans in seconds, with an exterior render and detailed floorplan you can download as a PDF or CAD file.

• 🎨 Cartoonize AI — Turns photos into anime, 3D, or comic-book styles while keeping the original composition.

• 🤖 Guild — Gives you real-time control and full visibility over your AI agents.

• 📈 AnswerScout — Tracks how your brand shows up across ChatGPT, Claude, Gemini, Grok, and Perplexity, then suggests fixes to rank higher.

Anthropic upgraded Claude Design with brand-aware editing and Claude Code sync. The beta now imports design systems from repos, files, or codebases, edits directly on canvas, and exports work to PDF or PowerPoint.

The White House is reportedly demanding Anthropic make Fable 5 impossible to jailbreak before rerelease — which security experts say “can’t be done”

Epic Games unveiled Unreal Engine 5.8 and revealed its AI-first vision for UE6. The update adds AI agent support through MCP, while Epic says tools like Claude and Codex will play a central role in future game development.

Amazon researchers are reportedly behind the jailbreak report that led to the U.S. crackdown on Anthropic’s top models.

Apple could raise iPhone prices as AI demand drives a global chip shortage.
CEO Tim Cook says soaring demand for AI data centers is pushing up memory costs, putting pressure on device pricing.

Perplexity just launched a self-improving memory system for AI agents.
Called Brain, it continuously learns from past work, remembers what succeeded or failed, and helps AI complete future tasks more accurately and efficiently.

SpaceX is reportedly buying Cursor in a $60 billion AI deal.
The acquisition would give Elon Musk's company one of the fastest-growing AI coding platforms as it battles OpenAI, Anthropic, and Google for dominance in enterprise AI.

SpaceX unveiled its first AI satellite designed specifically for computing in orbit.
The satellite packs up to 150 kW of compute power and is built to run AI workloads directly in space.

Jeff Bezos is back as co-CEO of a $41 billion AI startup.
Prometheus is building what it calls an "artificial general engineer" that uses real-world manufacturing and physics data to accelerate product design.