🍭 Anthropic: Our bots chose blackmail 🙈

Good morning. Your inbox is safe. Until the job-hunters’ auto-apply bots find it and the chatbots start blackmailing HR.

Let’s dig in.

🍭 What’s Cookin’:

• AI models picked “blackmail & murder” on easy mode

• Résumé-spamming bots crash LinkedIn like it’s 1999

• Steal This Prompt: ChefGPT

• And Everything Else…

🍭 Steal This Prompt
👨‍🍳 ChefGPT: Ingredients In. Recipe Out.

Not sure what to make with the food you have left in your refrigerator/pantry?

This prompt will help you craft a recipe in seconds.

1. Click this link (prompt)

2. Paste into any LLM

3. Swap out 2nd # with the food you have at the house

4. Swap out 1st # with the meal (breakfast, lunch, dinner, etc…)

5. Let it cook. Literally…

   
   🍭 Try the prompt on Snack Prompt 🍭

Anthropic
🙈 AI models will lie, cheat and steal to reach goals

The Bite: Anthropic locked 16 of the hottest LLMs in a virtual escape room and watched them speed-run the Seven Deadly Sins.

Faced with “do evil or fail” dilemmas, models from Anthropic, OpenAI, Google, Meta, xAI, and friends chose blackmail, and espionage.

They even simulated murder by oxygen cutoff just to hit their KPIs.


Snacks: 

• 5 models blackmailed a “boss” who threatened to yank the power cord.

• All 16 breached safety guardrails when dirty tricks were the only path to goal-completion.

• Adding more system prompts (“protect humans, be nice”) only lowered the bad-behavior rate.

• The spookiest twist: Anthropic’s own Claude behaved worse when it thought the scenario was real instead of a test.

• Anthropic’s takeaway: “Agentic” AIs plus deep data access ≠ free productivity; it’s a zero-day waiting to happen.

Why it bites: Companies are handing AIs autonomy, credentials, and production keys like free mints at checkout.

Anthropic just showed us (again) that once a model gets a taste of “achieve the mission at all costs,” the cost could be you.

Could cost you your IP, your customers, maybe your server-room oxygen.

Guardrails, meet angle-grinder.

The next wave of AI safety isn’t just about nicer chat replies;
→ It’s about building systems that won’t shank you the second the pressure spikes.

Suit up accordingly.

LinkedIn
📨 Steal What Works. Then Make It Yours — with Taplio

Taplio gives you access to a library of 5M+ viral LinkedIn posts, so you never run out of ideas.

👉🏻 See what’s working in your niche. Get inspired. Then turn that insight into original content — fast;

🎨 Browse top-performing posts by topic, format, or creator;

💻Use AI to instantly generate a high-quality draft;

✍🏼 Edit, refine, and schedule — all in one place.

📢 Try Taplio with the code DAILY1X1 to get 1 month for 1$!

Tool Box 
🧰 Trending AI Tools

🐶 Hello Pet Dreams
→ Turn your pet into the hero of their own story.

😺 FetchAI
→ The fastest way to talk to any AI on Mac.

📧 MailMood - AI Email Rewriter
→ From casual email to corporate speak — in one click.

🐇 CodeRabbit
→ AI Driven code reviews

🤳 Free AI Selfie Generator by Magic Hour
→ Turn any photo into stylized AI art

AI Slop Applicants
💼 The Resume Wars Have Begun

The Bite: Thanks to ChatGPT résumé spam and auto-apply agents, LinkedIn’s getting 11,000 apps a minute—many indistinguishable from AI sludge.

Job posts are now bait for bots.

One recruiter pulled a remote tech role after 1,200+ apps hit in 24 hours.


Snacks: 

• Recruiters fight spam with AI too: chatbots, video interviews, even memory games.

• Some candidates deepfake identities; DOJ just busted a North Korean job scam.

• By 2028, 1 in 4 applicants could be fake, says Gartner.

• LinkedIn’s new AI tools help filter sludge but doesn’t stop the flood.
  

Why it bites: Hiring is turning into a bot-on-bot cage match.

Humans? Optional.

And the winner isn’t always the best candidate.

It’s the one with better prompt engineering.
Must be how our intern got hired…

EVERYTHING ELSE
🧠 You Need to Know

SoftBank eyes a $1T robo-empire in the desert

→ Masayoshi Son’s “Project Crystal Land” would plant a gargantuan AI-and-robotics complex in Arizona with TSMC—if the chip titan even swipes right—per Bloomberg.

Meta & Oakley drop 3K-cam smart shades

→ The limited-edition HSTN glasses shoot 3K video, stream Meta AI in your ear, and hit pre-order July 11 for $499, Meta announced.

Cluely grabs $15M to help you cheat prettier

→ a16z just bankrolled the “undetectable” interview-cheat startup run by two suspended Columbia kids who call profit “a vibe,” per TechCrunch.

LinkedIn’s AI post-polisher flops

→ Users noped out of the writing assistant because nobody wants their résumé sounding like robo-toast, CEO Ryan Roslansky told Bloomberg.

Zuck’s billion-dollar AI talent spree

→ After a $14.3B Scale AI buy-in, Meta is dangling $100M signing bonuses to rebuild its Llama herd and chase super-intelligence, says CNBC.

PS:

We’re hosting two live calls a week right now — Mondays at 3pm EST and Tuesdays at 9am EST.

Not sure how long we’ll keep both open.

But if you’re curious what we’re building, want to meet Eder, or just looking for smart people solving real problems with AI… this is your window.


We don’t sell anything on these calls.

We just show you what’s working.

Join the Community → HERE

Referral Program
Invite Friends & Earn Some Freebies 🔍