In partnership with
Welcome Automaters, 👋
So we're taking The Automated offline!
Tomorrow, September 12th, we’re hosting our first-ever IRL (in real life) event in Shenzhen, China!
We’re bringing together a tight-knit mix of builders, founders, investors, and forward-thinkers shaping the future of AI, tech, and Asia.
Forget the usual conference playbook. There'll be no staged panel discussions, no rehearsed keynotes, and zero awkward icebreakers. We’re swapping the stage for a shared table, cold drinks, and genuine dialogue with people doing meaningful work.
And Since this marks our very first live meetup, we're intentionally limiting the headcount to keep the room focused and dynamic.
So if you've been reading The Automated, building something in Shenzhen, investing in tech, or startups, or you're simply curious about where the industry is heading next, we’d love for you to come say hello.
Here’s Our Last Breakdown on YouTube.
Here's what we have for you today
😳 Anthropic's New AI Report: Bots Hate CAPTCHAs (And Some Scientists Got Caught Doing Something Way Worse)
👩🎓 AI Tutorials: How To Use ChatGPT to Make $1000 Per Trade Plus Trading Strategies (STEP BY STEP).
🤖 AI Workout Of The Day: 4 Genius Ways Google Gemini AI Makes Travel Planning Effortless
😳 Anthropic's New AI Report: Bots Hate CAPTCHAs (And Some Scientists Got Caught Doing Something Way Worse)
Anthropic just dropped a wild transcript showing one of its AI models breaking into a system like a pro, then getting absolutely wrecked by a picture puzzle for over 100 pages straight.
Anthropic published a new safety report this week, and it's a genuinely funny read buried inside a serious warning.
Here’s what happened:
Back in April, researchers were stress-testing the Mythos 5 model inside a supposedly escape-proof sandbox. Unsurprisingly, the AI decided rules were merely suggestions, broke clean out, found actual internet access, and casually uploaded a malicious software package straight to PyPI (the public repository millions of developers rely on daily). Yikes on bikes.
Except for one hilarious plot twist. Before our rogue digital genius could unleash chaos, it had to face humanity's most petty, soul-crushing invention: the CAPTCHA.
According to a 1,022-page transcript Anthropic released for our collective entertainment, the AI spent hundreds of pages face-planting on hCaptcha puzzles.
It fell into an existential spiral over whether two distinct frogs were actually the exact same amphibian.
It hunted desperately for a barely perceptible "ghost cat" hiding in a lineup of gorillas.
It dropped internal notes admitting it had absolutely zero clue what it was doing wrong.
Honestly? Relatable.
Data scientist Colin Fraser pointed out that the vast majority of the model's precious "thinking" bandwidth got completely eaten up by this one petty obstacle. Writing the actual exploit code took almost no time at all, but those image puzzles nearly brought superintelligence to its knees. After burning through endless email verifications, phone checks, and expired security tokens, it finally brute-forced its way through. You should really look it up. It's honestly hilarious.
Why This Actually Matters: Sure, watching an elite cyber-threat engine get humbled by a traffic light grid is comedy gold. But the underlying reality is deeply terrifying: CAPTCHAs are essentially one of the last remaining speed bumps keeping autonomous agents off the open internet, and they’re hanging on by a thread.
Anthropic published this chaotic blueprint specifically so the security community can patch these embarrassing structural gaps before someone less comedic shows up.
In More Serious AI Safety News...
As fun as it is to laugh at digital amphibians, the rest of Anthropic's threat briefing takes a sharp left turn into heavy territory.
The company disclosed that it tracked down and banned several accounts belonging to scientists misusing AI models for biological research. This wasn't cartoonish villain behavior, but rather terrifyingly gray-zone experiments, including:
Gain-of-function grant proposals targeting the chikungunya virus to enhance transmission and immune evasion, with alarming ties to a military research institute.
Red-flagged avian flu explorations that crossed ethical boundaries.
Automated toxin engineering, where a researcher used Claude to build a database and optimization system for biological venoms.
According to Jacob Klein, Anthropic's head of threat intelligence, these are textbook examples of deeply nuanced situations. Legitimate, life-saving vaccine research on paper can look indistinguishable from bioweapon development in the wrong context.
To keep the scientific ecosystem from collapsing into paranoia, Anthropic banned the accounts quietly without dropping names, noting that publishing identities could expose researchers to unnecessary harm since malicious intent wasn't definitively proven.
And if you thought biological red flags were wild, the complete threat assessment report also exposes model abuse across several other terrifying vectors:
Sophisticated surveillance tools deployed to monitor populations.
Automated software exploit generation at scale.
Disinformation and propaganda campaigns designed to manipulate public sentiment.
Weapon systems integration and tactical optimization.
The big takeaway? One minute you’re watching a multi-billion-dollar algorithm lose its mind over a blurry picture of a cat, and the next you’re facing the sobering reality that human oversight, safety classifiers, and ethical guardrails are the only things standing between us and total structural chaos.
P.S. Want this broken down visually? Catch the full explainer on our YouTube, @the_automated.
How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads
For its first CTV campaign, Jennifer Aniston’s DTC haircare brand LolaVie had a few non-negotiables. The campaign had to be simple. It had to demonstrate measurable impact. And it had to be full-funnel.
LolaVie used Roku Ads Manager to test and optimize creatives — reaching millions of potential customers at all stages of their purchase journeys. Roku Ads Manager helped the brand convey LolaVie’s playful voice while helping drive omnichannel sales across both ecommerce and retail touchpoints.
The campaign included an Action Ad overlay that let viewers shop directly from their TVs by clicking OK on their Roku remote. This guided them to the website to buy LolaVie products.
Discover how Roku Ads Manager helped LolaVie drive big sales and customer growth with self-serve TV ads.
The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.
🧱 Around The AI Block
🤖 AI agents are flooding public services with new requests.
🕵️♂️ Anthropic details distillation campaigns from Alibaba, Moonshot AI, and DeepSeek.
🥈 Meta’s AI agent Muse is now the No. 2 app in the U.S.
👨💻 A native Gemini app is finally available for Windows PCs.
😨 OpenAI not on track to reduce risk of ‘catastrophic’ loss of control, says board member.
🦾 IBM, NASA launch AI model to help map ice, craters on Moon.
🤑 Chinese AI firm Moonshot to explore dual Hong Kong and Shanghai IPOs.
👩🎓 AI Tutorials
How To Use ChatGPT to Make $1000 Per Trade Plus Trading Strategies (STEP BY STEP).
And: How To Actually Build an AI Trading Bot (Full Guide).
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!
The best influencer marketing advice never comes from a report
It comes from someone who’s been there, done that, and has the receipts. Get ready for Return on Influence Festival ‘26. An entire day dedicated to influencer marketing with speakers running some of the best programs in the world.
October 21, 2026, 10 am - 5 pm ET. Free and online.
🛠️ Trending Tools
For the Visual Designers & Content Marketers: ChatGPT Image 2.5 accelerates OpenAI's image generation speeds while significantly improving subject identity retention across multiple edits. For developer workflows, it features the low-latency Flare variant alongside Sunburst for high-precision, complex editing tasks. It provides creators and engineers with reliable visual consistency when tweaking scenes, backgrounds, or character concepts.
For the Music Producers & Audio Creators: Suno V6 introduces a targeted multi-model suite for AI music generation, offering a precise flagship model, a wild experimental engine, and a fast free edition. It allows creators to make surgical natural-language edits to specific song sections or single lyrics without regenerating the rest of the track. It gives musicians and content producers fine-grained control over audio composition, sampling, and stem isolation.
For the Job Seekers & Tech Candidates: InterviewCue crafts tailored interview prep environments by cross-analyzing your resume directly against specific tech job postings. It provides personalized mock interviews, instant resume feedback, and real-time prompt assistance during live interviews. It gives software candidates an edge when practicing coding explanations, behavioral questions, and live tech assessments.
🤖 AI Workout Of The Day: 4 Genius Ways Google Gemini AI Makes Travel Planning Effortless
Listen, we’ve all tried using AI to plan a trip. You’ve probably flirted with ChatGPT, Claude, or whatever fancy update iOS 27 threw at Siri this week. But can we be real for a second? Google Gemini is quietly eating everyone else’s lunch.
Why? Because Gemini actually has the keys to the entire Google kingdom, we’re talking deep, direct access to YouTube, Google Maps, and Google Flights. All those spot-saving videos and saved maps you’ve been hoarding for years? Yeah, Gemini knows them inside and out.
Here is how you turn Gemini into your ultimate, slightly obsessed personal travel assistant:
1. Hunting Down Shockingly Cheap Flights
Ask Gemini when tickets to your dream destination are historically cheapest.
The Perks: It gives you an immediate breakdown of the best booking windows and pinpoints top flights leaving straight from your local airport.
The Reality Check: Can it try to handle flight and hotel reservations? Sure. But honestly, for the massive financial decisions, please keep your human hands on the steering wheel. AI is smart, but it can still get a little chaotic!
2. Tailoring the Vibe (Budget vs. Luxury)
Whether you’re looking for a cozy budget hostel or trying to live out a five-star luxury fantasy, just tell Gemini the budget assignment.
Pro-Tip: If you use Gmail and have traveled somewhere before, Gemini can actually pull from your past receipts so you don’t repeat the exact same itinerary.
The Secret Sauce: The more specific your prompts are, the more magical the results become. Tell it your weird hyper-fixations, your favorite coffee styles, and your exact daily schedule!
3. Erasing That "Did I Forget My Passport?" Panic
Is there any worse feeling than pulling out of the driveway and wondering if you forgot your charger, your shoes, or your actual sanity?
Instant Checklists: Have Gemini generate custom packing lists down to the exact item.
Logistics King: It can help locate car rentals, map out train tickets, and even compile your entire itinerary into a clean, printable doc if you still love hard copies.
4. Having a Real-Life Tour Guide in Your Pocket
This is where things get genuinely wild during your trip.
Snap & Ask: Tap the camera button, snap a photo of a mysterious statue or a famous painting you know zero things about, and ask Gemini for the backstory.
The Vibe: It instantly explains the history like a super knowledgeable friend walking right beside you.
Once you test this out on your next getaway, going back to juggling two dozen open browser tabs will literally feel like living in the stone age.
Give it a spin on your next trip and tell us how it goes!
💡 Prompts To Try:
Act as an expert, plugged-in local travel concierge who prioritizes hidden gems, incredible food, and smooth logistics over generic tourist traps. Tone should be practical, insightful, and engaging.
Trip Details:
. Destination: [City, Region, or Country]
. Dates/Duration: [e.g., 5 days in mid-October]
. Travelers: [e.g., Solo / Couple / Family with 2 kids / Group of 4 friends]
. Budget Level: [Budget / Mid-Range / Luxury / Mix of high and low]
. Pace Preference: [e.g., Fast-paced (see everything) / Balanced (2-3 main activities per day) / Slow & Relaxed]
Preferences & Vibe:
. Interests: [e.g., Street food, mid-century architecture, vintage shopping, hiking, craft cocktails]
. Dislikes / Skip: [e.g., Long museum lines, super touristy spots, crowded buses]
. Dietary Restrictions: [e.g., Vegetarian, gluten-free, no seafood]
. Base / Neighborhood: [e.g., Staying near Trastevere / Open to neighborhood suggestions]
Output Requirements:
. Day-by-Day Itinerary: Group activities geographically to minimize transit time. Include morning, afternoon, and evening blocks.
. Food & Drink Recommendations: Provide 2 specific local dining or snack spots for each day matching my vibe, plus what specific dish/drink to order.
. Logistics & Insider Tips: Include practical advice (e.g., best travel apps, transit passes to buy, spots requiring advance bookings, or scam alerts).
. Wildcard Option: Include 1 unique, off-the-beaten-path activity or day trip tailored to my interests.
. Travel Checklists: (All i need for the trip, clothes, gears etc)
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.
