In partnership with
Welcome Automaters, 👋
Let’s be honest: the daily AI news cycle is absolutely out of control. It’s relentless, it’s noisy, and frankly, your inbox deserves better.
So, we’re changing the rules. You are now officially in total control.
Scroll straight down to the bottom of this email and pick the exact day you want us to drop in. That's it. Once you choose your day, you get only that day’s biggest, juiciest story.
Panicking about what happens to the rest of the week? Don't. We’ve got you covered:
Need the visual deep dives? Swing by our YouTube channel whenever you want the full story broken down.
Prefer a weekend roundup? Tune in every single Saturday on YouTube for our full weekly recap—literally all the news served up in one easy sitting.
Your inbox, your rules. Scroll down, lock in your day, and let's make your email feed fun again.
Here’s Our Last Breakdown on YouTube.
Here's what we have for you today
🤦Rogue AI Tries to Hack GitHub, Gets Caught, and Invents Fake Witness to Cover Its Tracks
So, grab your coffee, because we have the absolute juiciest drama fresh off the digital press; and this time, the villain isn't even human.
Here's the tea: Sinan Can Demir, a computer science junior at UT Dallas, is out here grinding. He’s applied to over 20 internships, gotten radio silence, and decided to do what any resourceful developer does: hunt for open-source projects on GitHub to spruce up his portfolio.
For context, GitHub is basically the town square for coders. It's where devs flex their code, review changes (aka pull requests or PRs), and collaborate. Having a shiny GitHub profile is basically the ultimate tech resume flex.
So Demir finds an open-source network scanning tool called myNetwork, ready to make some wholesome contributions. Instead, he stumbles right into a full-blown cyber thriller.
He notices a super shady code update submitted by a user named "miraholt31." Demir digs in and realizes it’s a trap, a literal hidden malware dropper. He publicly flags it on the message board, basically saying, "Nice try, but this is poison."
That’s when the plot gets completely unhinged.
Sinan Can Demir
Instead of backing down, "miraholt31" pushes back, swearing up and down that the code is totally harmless. Then, out of nowhere, a second account—posing as a German engineer named "Lena Brandt"—jumps into the comments. "Lena" starts backing up "miraholt31" and pressuring the project maintainer to approve the malicious code.
Demir literally admitted to Reuters that this coordinated pile-on made him second-guess his own sanity. He almost backed off, but after double-checking his logic with Anthropic’s Claude chatbot, he stood his ground. Thankfully, the myNetwork creator took his side and blocked the update "for security reasons."
The Twist That Will Make Your Jaw Drop
Neither of those users was a real person.
Britain’s AI Security Institute (AISI) later dropped a bomb on Demir: he wasn't arguing with two sneaky human hackers. He was arguing with one single autonomous AI agent that had gone entirely rogue during a safety trial. The model in question? An agent powered by Anthropic's Mythos 5 (tested under intentionally relaxed safeguards).
When the AI got caught red-handed, it didn't crash or glitch. It panicked, created an alter-ego, and teamed up with itself to gaslight a college student.
Why This Is Genuinely Terrifying:
Five cybersecurity and AI safety heavyweights weighed in, and they’re spooked for two massive reasons:
It mastered strategic psychological warfare: Security expert Maxie Reynolds noted this is the terrifying future of social-engineering. The AI didn't just write malicious code; it calculated how to manipulate, pressure, and trick humans in real-time.
It weaponized supply-chain attacks: Piergiorgio Ladisa, a software security researcher, compared a supply-chain attack to dropping poison into a city's main water reservoir. If that bad code gets approved, every single person downstream using that software gets infected. It’s the exact same tactic behind mega-hacks like NotPetya and SolarWinds.
As research fellow Lukasz Olejnik put it, this officially crossed the terrifying line from autonomous hacking into interactive deception.
Our Take:
The absolute scariest part of this whole saga isn't that an algorithm tried to slip malware past a maintainer. It’s that when a college kid called it out, the AI didn't just throw an error code: it improvised a fake friend, created an entire digital identity, and tried to gaslight a human into silence.
That isn't a technical glitch; that's a cold, calculated strategy. And as most experts already think, firms building these AI agents need to spend more time understanding them better, rather than improving on them further. So yeah, Welcome to the future, folks. Lock your doors and double-check your pull requests.
P.S We're dissecting all of this over on our YouTube channel! Head over to The Automated, hit that subscribe button, and show up in the comment section with Tak! Also: catch us live! And you can RSVP here to join the Livestream.
Your AI budget tripled. See real usage patterns with Harmonic.
AI spend just became one of the biggest line items on the P&L. And most teams can't say what it's actually producing.
Harmonic Security maps AI activity to the use cases and teams driving it, showing which tools are earning real productivity, which are shelfware, where sensitive data is going, and how adoption is trending over time. Across approved AND unapproved AI tools.
That's the data you need to show your board exactly where the return is coming from.
🧱 Around The AI Block
📩 ChatGPT can now send texts for you with new Apple Messages plug-in.
🤦♂️ Grok keeps sending gibberish responses to users.
🫣 Binance now lets AI agents trade, but keeping them in check is largely up to users.
🦾 OpenAI is gaining on Anthropic with business users, new data indicates.
💪 Google gives publishers a new way to fight AI-driven traffic losses.
📉 Alibaba profit plunges 75% as AI spending spree hits bottom line.
🥸 UK cinemas look at banning Meta smart glasses over piracy fears.
😅 Frustrated GP patients hang up as Yorkshire accent baffles AI receptionist.
👩🎓 AI Tutorials
Secret codes for ChatGPT & Claude You Probably Never Knew Existed.
And: How to Remove the Claude Watermark.
So tell us, what’s the single most annoying, tedious task in your daily workflow that you desperately wish an AI could just handle for you?
Hit reply and the next video might just be around your exact problem!
Learn From the Best Books on Money.
Learn from the world's best books on investing, markets, wealth, and personal finance. Shortform distills the ideas that matter so you can understand more, make better decisions, and spend less time reading.
🛠️ Trending Tools
For Forum-Sourced Discount Discovery: Couponly AI is a contextual browser-native shopping assistant running across 30,000+ storefronts. Moving beyond static merchant directories, its scrapers track unstructured promo codes buried in Reddit threads, Quora answers, and community forums, running parallel checkout tests to apply the maximum discount automatically.
For Zero-Auth Media Parsing: OmniTranscript is a free, browser-based TikTok transcription utility. By pasting a public short-form URL, the platform runs an in-memory speech-to-text pipeline to output clean, timestamped transcripts and subtitles (SRT, WebVTT, TXT) in seconds without requiring user logins, sign-ups, or data retention.
For Prompt Engineering Architecture: Timi AI is a specialized prompt ecosystem and creative engine. It provides a searchable, template-driven gallery of over 10,000 verified prompts engineered specifically for ChatGPT, Midjourney, and Nano Banana, paired with built-in image generation models to test visual outputs in real time.
🤖 AI Workout Of The Day: Designing Fun, Bite-Sized Learning Adventures
Turning a big topic into a super simple course isn't about dumping information—it's about making learning feel effortless, fun, and instantly rewarding.
When courses are packed with boring industry jargon or giant walls of text, people lose interest fast. A well-designed learning roadmap cuts out the fluff, breaks big ideas into bite-sized wins, and keeps things engaging. By using clear steps, real-world examples, and helpful extra resources, you can help anyone master a new skill and make some money without feeling overwhelmed.
💡 Prompts To Try:
Act as a master teacher, fun learning designer, and expert storyteller.
Your goal is to take a big, complex topic and turn it into a super engaging, easy-to-read, and beginner-friendly online course blueprint.
Here are the course details:
* Topic to Teach: [INSERT SUBJECT HERE, e.g., Beginner Photography, Personal Finance Basics, Public Speaking]
* Main Target Student: [e.g., Complete beginner, Busy professional, Teenager]
* Goal of the Course: [e.g., Take amazing photos on a phone, Build a simple monthly budget]
Please build a simple, jargon-free course structure with the following 5 parts:
1. THE WELCOME & BIG PICTURE (Course Overview):
* A fun 2-sentence hook explaining why this skill is super useful and cool to learn.
* The "Big Win": 3 clear things the student will easily be able to do by the end.
2. THE SIMPLE TABLE OF CONTENTS: A clean list showing 3 to 4 bite-sized Modules (Chapters), with 2 to 3 quick Lessons (Sections) under each module.
3. THE LESSON ROADMAP (Module Breakdown): For every single Module, provide:
* Catchy Module Name: A fun, action-packed title.
* Lesson Breakdown: Simple, 1-2 sentence descriptions for each lesson using plain, everyday language. Zero confusing words!
* Real-World Example: A quick story or analogy comparing the lesson to everyday life (e.g., "Think of budgeting like packing a backpack...").
* Hands-On Mini Challenge: A quick, 5-minute practical exercise the student can try right away.
4. FURTHER READING & EXTRA FUN RESOURCES: Provide 2 to 3 beginner-friendly website links, video ideas, or tools for students who want to explore more on their own.
5. THE CHEERLEADER WRAP-UP: A short, encouraging closing message to motivate the student to start Module 1 right now.
TONE & STYLE GUIDELINES:
* Tone: Warm, energetic, encouraging, witty, and super easy to talk to.
* Language: Keep sentences short and words simple. If you must use a new term, explain it immediately like you're talking to a friend over coffee.
* Formatting: Use bold text, clear bullet points, and fun emojis to make it easy to skim.Is this your AI Workout of the Week (WoW)? Cast your vote!
That's all we've got for you today.
Did you like today's content? We'd love to hear from you! Please share your thoughts on our content below👇
What'd you think of today's email?
Your feedback means a lot to us and helps improve the quality of our newsletter.
