Welcome back! Google’s new AI can trace a disease to the exact cells where it starts. Anthropic says Claude moved itself onto AMD’s chips in just three days. Alibaba gave the world a first look at Qwen4, but only for two weeks. And Amazon’s latest release comes with a surprise about who built it.
In today's Generative AI Newsletter:
Health: Where does a disease really begin?
Anthropic: How did Claude move itself to new chips in three days?
China: Why does China’s newest AI expire in two weeks?
Amazon: What’s the surprise in the latest AWS release?
Google’s AI spots where disease begins
Google DeepMind released AlphaGenome 2, an AI that predicts how a change in your DNA affects the genes inside one specific type of cell.
And it finds nearly twice as many of the changes that matter as the best model before it.
The first AlphaGenome looked at whole tissues. This one zooms in on a single cell type, which helps scientists find the exact cells that drive a disease.
100 million+ individual cells from 68 tissues were used to train it.
30% of the time it found the DNA change causing an effect in immune cells. Other tools, including the first AlphaGenome, managed 16%.
1 DNA change behind inflammatory bowel disease was pinned down to the gut immune cells where it does its damage.
1 change linked to severe COVID-19 turned out to disrupt genes more strongly in activated immune cells.
0 dollars for non-commercial research. Scientists can use it now through an API.
Could AI find the cells behind a disease before doctors do?
Claude ported itself to AMD chips in three days
Anthropic revealed that Claude Mythos, its most powerful model, moved Claude onto AMD’s chips in three days. It shared the numbers in its IPO roadshow, the pitch it’s giving investors before its stock market debut next week.
And it did almost all of the work itself.
Moving an AI to a new kind of chip means rewriting the low-level code that tells the chip what to do. Here’s how Anthropic says it went:
3 days from start to Claude running on AMD, with “almost no human intervention.”
Nearly 100% of that code was written or moved over by Mythos.
6 to 12 months is how long the same job used to keep dozens of engineers busy.
4 kinds of chips now run Claude: Amazon’s, Google’s, Nvidia’s and AMD’s.
1 “unconventional bet” is how co-founder Tom Brown describes running on so many chips, one most of the industry thought wasn’t worth it.
AMD is also an Anthropic investor, and Anthropic plans to run much more of Claude on its chips starting next year.
Months of engineering, done in three days.
Alibaba’s Qwen4 arrived with an expiry date
Alibaba opened a preview of Qwen4-Max, the top model of its next AI family. Its message was three words: “Qwen4 is coming.”
And the preview disappears in two weeks.
Alibaba promises “deeper knowledge,” “richer world understanding” and “stronger coding and agentic capabilities.”
1 million tokens of context, enough to read a whole codebase at once.
3 kinds of input: text, images and video.
14 days is how long the preview lasts, from November 11 to November 25. After that, the preview model is retired.
1 way to pay for it: Alibaba’s Token Plan subscription. There’s no pay-as-you-go option yet.
0 benchmark scores have been published, and there’s no date for the full release.
You can also try it early in the Qwen app, QwenStudio, QwenWork and Qoder.
Amazon’s latest release has a surprise
Amazon released AWS Emulator, a free tool that lets developers test their apps on their own computer instead of Amazon’s cloud. Every line of its code was written by AI agents.
And no one on the team has reviewed any of it.
“Our team hasn’t written or reviewed any of the code,” AWS senior principal engineer Matthew Goldberg wrote. It started as a weekend experiment:
15 minutes was all Amazon’s AI coding tool, Kiro, needed to build a first working piece. It built the rest over the weekend.
3 teams of AI agents now keep it running: one writes tests, one finds bugs and one fixes them.
Hundreds of agents fix thousands of issues every day, Amazon says.
1 checkpoint of AI reviewers inspects every change for quality and security before it goes in.
It’s free and open to anyone on GitHub.
Amazon says the idea came from Anthropic’s experiment with an AI-built C compiler.
Amazon’s engineers built the agents. The agents built the product.
Tool of the Day: Lindy
Lindy is an AI assistant that works inside the apps you already use, like your email, Slack and calendar, and takes real work off your plate. Ask it in plain words to chase an overdue invoice, reply to new leads after hours or send a client report every Monday, and it does the job. It learns how you work from your emails and calls, and anything that goes outside your company waits for your OK first. You can start free with no credit card.
Try this yourself:
Sign up free at lindy.ai for $50 in credits, good for 7 days.
Connect Gmail, Slack or your calendar.
Give it a task, like “Chase every unpaid invoice.”
Approve what it sends.
Who it's for: freelancers, small business owners and anyone buried in follow-ups and admin work.
Everything else you shouldn't miss
Experts confirmed OpenAI’s AI-written proof of a famous physics problem: a committee led by Nobel laureate David Gross says it proves Yang-Mills theory exists, half of a $1 million Millennium Prize problem, though they admit they “do not claim to fully understand the proof yet.”
Russian pranksters fooled OpenAI’s national-security team into a video call: posing as Ukraine’s spy chief, they kept three senior staff talking for 17 minutes, and OpenAI confirmed it was a prank.
Grok Build left more than 35,000 apps open to anyone: a security firm found medical exam results, employee IDs and 24 of SpaceXSI’s own internal apps exposed through share links.
United won’t let your AI agent book your flight: agents can still show you flights, but United’s CEO says “I don’t want to create another middleman.”
Gemini is now on Apple CarPlay: iPhone users can talk to Gemini hands-free on their car screen to catch up on email, find places or rehearse a presentation.
Learn more about AI from the experts building it
📸 Follow us on Instagram for fast, visual AI updates in 30 seconds.
📺 Watch us on YouTube to hear insights directly from leading AI voices, builders, and innovators.
🐦 Follow us on X for breaking AI news and real-time industry updates.
🧠 Learn how to build your next AI application with practical resources and expert guidance.
🎓 Check out courses made by AI experts on GenAI Academy.
💰 Invest in the GTM infrastructure of the AI economy.
