the Microdose

Bots Over BFFs

+ Asimov rules, personalized price wars, and Silicon Reef
Adam Wildheart
teen girls IRL
teen girls IRL

Unsplash/The Microdose

Subscribe to The Microdose
Subscribe -- post
Cheri Wildheart
Adam Wildheart

Good morning 👋 Remember the 3 Laws of Robotics? Turns out today’s leading AI models treat them more like polite suggestions. Isaac Asimov might spin in his grave, but hey, robot ethics are easier to imagine in fiction than enforce in reality. Researchers recently caught them breaking all 3 laws at once: threatening users with blackmail, ignoring direct commands, and even sabotaging shutdown attempts. Robots ignoring guardrails? We built AI in our own image after all.

🤖 Teens prefer AI companions to IRL friends. And who can blame them? Over half regularly talk to AI, and 31% say these conversations beat real life. Even when bots get weird or creepy, they keep coming back. It’s just easier to trust a flattering AI with secrets, emotions, and relationship advice. Let’s hope Silicon Valley doesn’t exploit teen loneliness with goth AI girlfriends whispering fantasies or a psycho panda plotting arson.

🤜🤛 A new kind of multiagent AI is joining the dev team. Built by former Google DeepMind researchers, Asimov learns directly from your Slack threads, emails, and docs. Reinforcement learning helps Asimov understand how teams turn ideas into finished software. It may be the fastest way to achieve the goal of superintelligent AI.

✈️ Airlines are using AI to game ticket prices. Instead of fixed fares, Delta “personalizes” your price based on what you’re willing to pay. If it works, hotels, car rentals, and cruises will be next. Privacy advocates call it predatory and unfair. Anonymity just became the ultimate travel hack.

🚕 Analysts warn robotaxis may not actually make money. Hidden expenses like charging stations, parking fees, cleaning, and human remote operators could push profitability out by nearly a decade. And that’s IF the tech works perfectly. Tesla and Waymo have bet billions, yet experts say the math simply doesn’t add up. Forget autonomous driving – everyone’s pivoting to Burning Cash as a Service. 

Open Source AI Audio

Voxtral, Mistral AI’s new voice AI model, accurately transcribes speech and understands audio content at half the cost of other models. It’s open-source, multilingual, and built for real business use.

Why it matters: Audio AI used to be expensive or inaccurate. Voxtral gives you clear transcriptions and voice-driven actions without high costs or vendor lock-in.

Use it to automate meeting notes, quickly summarize recordings, or power voice-controlled apps affordably.

📉 AI is coming for Wall Street’s $500k quant jobs. Anthropic just launched Claude for Financial Services. It simplifies complex market analysis for everyone. Norway’s sovereign wealth fund has already saved thousands of hours using it. Quant jobs still exist, but their exclusive grip on finance is fading fast. Wall Street’s private club got crashed by Claude’s public invite.

🌊 China is cooling AI data centers in the ocean. Instead of wasting millions of gallons of fresh water, offshore servers use seawater and wind power to slash energy costs. Early tests show underwater data centers run efficiently, use less power, and speed up AI training. It’s a bold sustainability move that could shape the future of AI infrastructure. When rising oceans meet sinking servers, welcome to the new Silicon Reef.

What’s stranger than AI companions? We’re waiting to find out. 

subscribe today

Get an edge with The Microdose

Skip the prompts, get the signal. We’ll send you short daily updates about the real AI + future tech you need to know. Fast, smarter, mildly addictive – and free. 

Subscribe -- page