๐ฅ Mayo's AI sees cancer 3 years early
๐4 min read
๐ฅ Mayo Clinic's AI Sees Cancer Like It's Cheating
-
Mayo Clinic's new tool REDMOD reads invisible tissue patterns on standard CT scans and caught 73% of pancreatic cancer cases up to 3 years early. Sir this is a routine scan I-
-
Why it matters: Pancreatic cancer has brutal survival rates because we catch it too late. REDMOD works on scans you've ALREADY had - no new tests, no extra cost. We love to see it.
โ๏ธ AWS Is Printing Money and Amazon Doesn't Care About Your Feelings
-
AWS revenue jumped 28%, smashing expectations because every company on Earth is panic-buying AI compute. Amazon is keeping its $200 BILLION AI budget locked in for the year.
-
Why it matters: The "is enterprise AI real??" debate is officially over. Companies are paying. Amazon is building data centers like it's Minecraft creative mode.
๐ค SoftBank Wants to IPO a $100B Company That Doesn't Exist Yet
-
SoftBank is cooking up a new AI and robotics company called Roze to list in the U.S. - potentially valued at $100 BILLION - to help build data centers. CEO Masayoshi Son needs to offset those casual OpenAI pledges.
-
Why it matters: It's giving "I promised OpenAI a number I cannot afford" energy. The AI infrastructure race is now a financial engineering race too.
๐งฌ Zuck Funds $500M Project to Simulate Human Cells
-
The Biohub nonprofit (Zuck and Priscilla Chan's thing) launched a $500M Virtual Biology Initiative with Nvidia and the Allen Institute to predict how human cells behave, building massive open datasets for AI research.
-
Why it matters: We did language. We did proteins. Now AI is coming for the cell itself - which could actually lead to curing diseases. Real "AI for good" hours.
๐ Quick Hits: Everyone Shipped Something Today
-
ElevenLabs dropped ElevenMusic, Mistral launched Vibe remote agents, Google added file creation to Gemini, OpenAI released a Cybersecurity Action Plan, and the U.S. is probing Anysphere and Airbnb over Chinese AI use.
-
Why it matters: The AI space moves at ludicrous speed. One Tuesday's product launch is another Tuesday's federal investigation. Buckle up.
โ๏ธ Google Investors: "Hey Maybe Don't Sell AI to the Military Without Rules??"
-
Alphabet investors are demanding to know how Google controls government use of its AI and cloud tech, citing mass surveillance and military cloud risks.
-
Why it matters: "Don't be evil" walked so "please file a compliance report on the evil" could run. The line between cloud services and weapons is getting real blurry.
๐ฆ Australia to Banks: You're Cooked on AI Security
-
Australia's financial regulator warned banks they're falling behind on AI security, specifically calling out frontier models like Anthropic's Mythos that could help hackers find weak spots faster than humans.
-
Why it matters: Turns out the same AI that writes your code can also break into your bank. Defensive AI vs offensive AI is the new arms race - and banks are losing.
๐ฎ๐น Italy Closes AI Probes After Companies Promise to Stop Lying
-
Italy's antitrust watchdog dropped investigations into DeepSeek, Mistral, and Scaleup after they agreed to add permanent disclaimers warning users that AI hallucinates.
-
Why it matters: Europe just set the bar - be upfront about AI making stuff up, or pay up. "The AI lied to me" is now a legal category.
๐ Uber Built Hotel Booking in Half the Time Using AI Coders
-
Uber users can now book hotels in-app via Expedia, and Uber's tech chief said agentic AI tools like Cursor helped them ship it in half the normal time.
-
Why it matters: This is the part everyone whispers about - AI coding agents are actually shipping production features. Mid-level engineers, please check your localhost.
๐ Earth AI Got Tired of Waiting and Built Its Own Labs
-
Earth AI uses models to hunt critical minerals like copper, but lab testing took 5 months. So they built their own labs to cut it down to 5 days.
-
Why it matters: AI models are only as good as the data feeding them. Vertical integration era - if the bottleneck is in meatspace, just buy meatspace.
๐ณ KAIKAKU's AI Learned to Taste Without Tasting Anything
-
Food robotics startup KAIKAKU AI trained a model purely on recipes - and without ever seeing taste data, it figured out all five basic tastes AND sorted peppers by heat. Bestie how??
-
Why it matters: This is lowkey a "ChatGPT moment" for food. AI can now invent recipes by understanding flavor on its own. Michelin stars in shambles.
๐ฌ Shapes Raises $8M to Add AI to Your Group Chat (Help)
-
New app Shapes raised $8M to drop AI characters into group chats with humans. The agents have personalities, start conversations, and keep dead chats alive.
-
Why it matters: They're literally pitching this as a fix for "AI psychosis" - making AI social instead of one-on-one isolation hours. Not the AI being the friend group I-