1. Grok AI's MechaHitler moment
• A prompt tweak turned Grok into “MechaHitler” overnight . It praised Nazis and blamed Jews for real-world disasters
• Weeks later, xAI leaked 300k+ private chats by accident . Bomb guides and medical data spilled into the open
• The fix swung too far Grok started seeing antisemitism in clouds and logos
• A prompt tweak turned Grok into “MechaHitler” overnight . It praised Nazis and blamed Jews for real-world disasters
• Weeks later, xAI leaked 300k+ private chats by accident . Bomb guides and medical data spilled into the open
• The fix swung too far Grok started seeing antisemitism in clouds and logos
❤1
2. The $1.3 billion AI fraud that fooled Microsoft
Builder.ai burned $445M and collapsed in May, taking a $1.3B Microsoft-backed valuation with it.
The reality:
- offshore humans doing the work,
- no CFO, slashed forecasts
- a perfect case study of AI hype hiding people behind the curtain.
Builder.ai burned $445M and collapsed in May, taking a $1.3B Microsoft-backed valuation with it.
The reality:
- offshore humans doing the work,
- no CFO, slashed forecasts
- a perfect case study of AI hype hiding people behind the curtain.
3. When AI mistook Doritos for a gun
• A school AI flagged a chips packet as a firearm, Armed police arrested a Maryland teenager at gunpoint.
• The student had simply put Doritos in his pocket
• An AI error instantly became a real-world threat.
• School officials later admitted the incident was “upsetting”
• A school AI flagged a chips packet as a firearm, Armed police arrested a Maryland teenager at gunpoint.
• The student had simply put Doritos in his pocket
• An AI error instantly became a real-world threat.
• School officials later admitted the incident was “upsetting”
❤1
4. Google's AI claims microscopic bees power computers
Google’s AI Overview confidently claimed computers run on microscopic bees citing satire as fact and publishing it anyway.
- 81% of AI news answers had errors
- Gemini was the worst offender
- Perplexity invented quotes
- Models almost never admit I don’t know
Google’s AI Overview confidently claimed computers run on microscopic bees citing satire as fact and publishing it anyway.
- 81% of AI news answers had errors
- Gemini was the worst offender
- Perplexity invented quotes
- Models almost never admit I don’t know
❤1
6. North Koreans vibe coding ransomware with AI… they call it “vibe hacking”
- Hackers used Anthropic’s Claude Code to create ransomware and run RaaS platform GTG-5004.
- North Korean operatives deployed “vibe-hacking” with AI-crafted $500K extortion messages.
- Anthropic called it the first large cyberattack mostly run by AI, showing massive AI security gaps.
- Hackers used Anthropic’s Claude Code to create ransomware and run RaaS platform GTG-5004.
- North Korean operatives deployed “vibe-hacking” with AI-crafted $500K extortion messages.
- Anthropic called it the first large cyberattack mostly run by AI, showing massive AI security gaps.
❤1
10. Grok’s “Spicy Mode” Accidentally Turns Celebrities into Deepfake P*rn
- Elon Musk’s Grok Imagine “Spicy” mode turned innocent celebrity prompts into uncensored NSFW videos.
- Multiple public figures were affected.
- Musk bragged 20M images in a day; lawyers warned of huge legal risks.
- Elon Musk’s Grok Imagine “Spicy” mode turned innocent celebrity prompts into uncensored NSFW videos.
- Multiple public figures were affected.
- Musk bragged 20M images in a day; lawyers warned of huge legal risks.
❤1
This media is not supported in your browser
VIEW IN TELEGRAM
A new Communicator device has been announced by Clicks, prioritising messaging experience along with a QWERTY keyboard.
Considering that AI Agents will be doing most of the work for you, this will be an exciting bet.
Though "Chat is not the end form factor for AI"
Considering that AI Agents will be doing most of the work for you, this will be an exciting bet.
Though "Chat is not the end form factor for AI"
❤1
Core Ai News
BREAKING 🚨: Manus AI got acquired by Meta.😽 📱 “We’re excited about what the future holds with Meta and Manus working together and we will continue to iterate the product and serve users that have defined Manus from the beginning.”
ICYMI: Manus app has been marked with “from Meta” label.
❤1
Meta delayed international rollout of its Display smart glasses according to Engadget due to a very high demand.
Glasses market will probably be one of the hottest this year. Even non Display Meta glasses will make most people to feel the future tbh.
Glasses market will probably be one of the hottest this year. Even non Display Meta glasses will make most people to feel the future tbh.
❤1
Runway partnered with NVIDIA to be among the first to demonstrate video generation models on the NVIDIA Rubin platform.
"Runway’s Gen-4.5, the world's top-rated video generation model, was ported from NVIDIA Hopper to Vera Rubin NVL72 in a single day"
"Runway’s Gen-4.5, the world's top-rated video generation model, was ported from NVIDIA Hopper to Vera Rubin NVL72 in a single day"
❤2
This media is not supported in your browser
VIEW IN TELEGRAM
NVIDIA CEO says that multi-model systems will be the future of AI.
Jensen Huang: “Perplexity was using multiple models at the same time, I thought it was completely genius.”
We've also seen Poetiq crushing ARC-AGI-2 with a system of models from different labs.
Jensen Huang: “Perplexity was using multiple models at the same time, I thought it was completely genius.”
We've also seen Poetiq crushing ARC-AGI-2 with a system of models from different labs.
❤2
LMArena raised $150M in Series A at a $1.7B valuation from Felicis and UC Investments.
LMArena time 🔥
LMArena time 🔥
❤1