4-min read
This Issue: Claude Code is making auto mode the default, while North Korean hackers are using AI for spear-phishing and researchers are testing products with billions of synthetic personas. Plus: AI finds hidden health risks in sleep studies, Visoid brings AI deeper into architectural design, and a Claude app-building tutorial.
Claude Code Is Making Auto Mode the Default
Anthropic is making auto mode the default in Claude Code for Pro, Max, and Team users starting August 14. Instead of repeatedly asking developers for permission, Claude Code will keep working on its own unless an action is considered irreversible, destructive, or aimed outside the user’s environment.
Image: AI-generated
Highlights
Auto mode becomes the default for Pro, Max, and Team plans on August 14
Claude Code can continue working without asking for approval at every step
In Anthropic’s test of 1,053 professionals, auto mode caught 89% of dangerous commands versus 13.6% for human reviewers
Claude Code users currently approve 97% of permission prompts
Teams using auto mode are shipping about 25% more pull requests
Enterprise, API, and cloud-platform users remain opt-in for now
Why it matters: Anthropic is making Claude Code more autonomous, reducing permission fatigue while relying more on built-in safeguards to catch risky actions.
Your AI stopped learning the day it shipped.
An off-the-shelf model doesn’t automatically learn from your production failures, operator corrections, or the work that makes your business different.
On August 11, Oumi will show the complete loop: Evaluate, Synthesize, Train, Deploy, Compound.
Join Manos Koukoumidis and the Oumi team at 10:00 AM PT / 1:00 PM ET for a live demonstration on a real task.
See how teams can turn what happens in production into an AI advantage that compounds.
Reserve your seat
AI from Cleveland Clinic and IBM analyzed nearly 10,000 sleep studies and identified five patient risk groups using signals such as brain activity, heart rhythm, breathing, and oxygen. The highest-risk group had about twice the five-year death risk of the lowest-risk group — a difference standard sleep-apnea scoring missed. The research suggests AI could help doctors get more useful health information from sleep data they already collect.
Want it weekly?
🔥Big news in AI
Image: AI-generated
🔹North Korean Hackers Employ AI in Spear-Phishing Attacks
A report by South Korean cybersecurity firm Genians revealed that North Korean state-backed hacking group Kimsuky is actively using AI to automate cyberattacks. By utilizing open-source tools like Ollama and GPT-4All to run LLMs offline, the threat group generates highly polished decoy documents for large-scale spear-phishing campaigns targeting military, diplomatic, and academic sectors.
🔹invideo Brings Seedance 2.5 to Agent Two for Hybrid Video Workflows
invideo announced that the Seedance 2.5 AI model is now live on invideo Agent Two, targeted at serious creative production workflows. Focused on hybrid shoots, the integration excels at preserving an actor’s original performance and camera movements while seamlessly blending generated environments and complex visual effects. It also offers high scene and motion adherence to maintain consistency across edits.
🔹Researchers Introduce MatrAIx for Large-Scale AI Persona Simulations
Researchers from institutions including Harvard and MIT introduced MatrAIx, a population-scale simulation platform designed to test AI systems and digital products using 8.3 billion synthetic persona profiles. Powered by models like Claude Opus 4.8 and GPT 5.5, the platform evaluates product changes—such as pricing shifts or UI updates—across four interactive environments (Survey, Chatbot, Web, and App) to model diverse human behaviors and preferences at scale without high manual evaluation costs.
Warmly Ran GTM With No Sales Team. Here's How.
That's what Warmly proved. They defined ICP, scored buying intent, and surfaced the right accounts before a human ever touched a lead. HubSpot noticed.
On August 12, Max and Keegan are rebuilding it live in HubSpot — and showing you how to replicate it this week. HubSpot Credits included when you join HubSpot for Startups.
Visoid Raises $2.5M to Bring AI Into Architectural Design
Oslo-based Visoid raised $2.5 million to expand its AI visualisation platform for architects. The software turns existing 3D architectural models into high-quality renders within minutes, letting architects explore and communicate design ideas while preserving the original design.
The bigger shift is bringing AI earlier into the design process, not just using it for final renders. Visoid lets architects test and iterate on concepts directly from their existing models, while the new funding will support more AI capabilities, hiring, and international expansion.
Build a Working App With Claude AI in 8 Steps (Even If You’ve Never Coded)
This is a tutorial, not a story. By the end, you’ll know how to take an idea, build it into a working app with Claude, and put it online for free where anyone can use it.
AI Insights. Real Growth. Higher GMV, Better Profits
The difference between growing stores and stagnant ones isn't more effort. It's better insights. StoreClaw analyzes your Shopify and Amazon data, surfaces your biggest growth opportunities, and helps you increase GMV while protecting profit. Start free with bonus tokens. No credit card required.
shared.image.missing_image
AI’s Fastest-Growing Skill Isn’t Technical
Critical-thinking enrollments jumped 185% among GenAI learners as workers increasingly learn to evaluate—not simply generate—AI output. This Sunday Special also covers Google’s 12 free agent-building labs, a $180,000 hackathon, practical courses, upcoming events, and current AI career opportunities.
Want it every Sunday?
New here? Get the next issue Thursday.
Join 76,000+ AI builders who read The AI Entrepreneurs every week
