Welcome back! Apple's new Siri finally showed up at WWDC, the brain inside it belongs to Google, and not every iPhone makes the cut. OpenAI filed to go public a week after Anthropic did. Apple also opened its apps to outside models, so Gemini and Claude can now live inside your iPhone and inside Xcode. And Anthropic explained, with receipts, why the same AI that writes your code can't pull a clean list of virus sequences.
In today's Generative AI Newsletter:
Apple Intelligence: Does your iPhone even make the cut for the new Siri?
OpenAI: What does a confidential filing say that Sam Altman won't?
Gemini in Xcode: Why did the most closed company in tech hand Google a key?
AI in biology: Why can AI ship your app but not pull a clean list of virus data?
After two years of waiting for a smart Siri, yesterday's announcement comes down to two things.
Apple took its time revamping how Siri looks, to make it the most intuitive on-phone AI experience.
They handed the functionality to Google, which makes Gemini the new brain behind the assistant.
Examples of what Apple AI can do right now:
Camera: Point it at a plate of food for the nutrition breakdown, or at a bill to split it and send everyone their share over Apple Cash.
Shortcuts: Say "let my partner know when I leave work and send them an ETA from live traffic" and it builds the automation.
Safari: Describe the browser extension you want and it builds it.
Passwords: It finds your weak logins and, once you approve, signs into each site to change them and save the new ones.
The most capable on-device Siri only runs on the iPhone Air and the iPhone 17 Pro, not even the regular iPhone 17. Apple first promised this Siri back in 2024, and the iPhones that were out then don't make the cut either.
This doesn’t mean your phone isn’t getting a smarter Siri. You’ll just miss out on a few things like Siri’s new voices and running a few AI tasks locally.
Here’s the Tl;dr. Apple paid $250 million to settle a lawsuit over the Siri features it showed two years ago and never shipped. Not to mention the number of upset users. After a long wait and a settlement later, it seems like Apple AI went from very behind to good enough.
Most AI demos never reach production. They stall in pilot, get shelved or take 3x longer than planned.
From Demo to Production is a 6-week cohort course for technical leaders. You get a complete framework to operationalize agentic AI systems, plus a Production Blueprint tailored to your organization that you can present to leadership and implement immediately.
Btw, Ari Joury, PhD, CEO, and AI Decision Scientist dropped a free spotlight video from his course. Click here and scroll down a bit to find it.
OpenAI confidentially filed for an IPO on Monday, a week after Anthropic did the same. The valuation is around $852 billion, and reporting points to a debut as soon as September, though OpenAI hasn't locked a date.
We told you yesterday that OpenAI is turning ChatGPT into a superapp to put its enterprise revenue in front of investors. This is the other half of that move.
Goldman Sachs and Morgan Stanley are running it, Morgan Stanley being the bank that led Facebook's IPO. Those are the same two leads on Anthropic's filing, which came in at $965 billion and added JPMorgan as a third. It all lands days before SpaceX starts trading.
Pretty soon you'll be able to buy a piece of the company that's made half its users nervous about their jobs.
Until now, building AI into an iPhone app mostly meant using Apple's own model, which trails the best out there. That changed yesterday. Developers can now drop Google's Gemini or Anthropic's Claude straight into an app through Apple's own tools, by swapping a single line of code.
Apple also put Gemini inside Xcode, the software people use to build Apple apps, where it works as an agent that writes and fixes code. The preview opens today, and a starter Gemini key is free from Google AI Studio.
Apple spent fifteen years keeping its platform shut. Making rival AI the easy default inside its own apps is Apple admitting its models can't win on their own, and handing Google and Anthropic a road to every iPhone.
Anthropic published research showing AI is great at coding and bad at biology.
Software was built for machines. Biology's data hides behind web pages built for humans, so an agent asked the same question three times gave different answers. 106, then 15, then 5. The right answer was 266.
A simple tool that pulls the data the same way each time fixed it, pushing accuracy to 99.7% and letting cheap models match expensive ones.
Anthropic has skin in the game now. It just bought a biotech startup for $400 million and launched a life-sciences version of Claude, so it's arguing the next breakthroughs come from fixing boring data plumbing, right as it starts selling that fix.
Nile is a free app for people with epilepsy who are tired of tracking seizures in their notes app. You log an aura or a seizure in a couple of taps, add your medications with a reminder for each dose and build a clean history you can hand your neurologist instead of reconstructing it from memory in the room.
Try this yourself:
Download Nile AI from the App Store or Google Play.
Log your first entry, an aura or a seizure, with the time and anything that might have set it off.
Add your medications and set a reminder for each dose.
Let the history build, then bring it to your next neurology appointment.
Who it's for: anyone managing epilepsy, or a caregiver tracking it for someone else, who needs the record to be right.
Data centers chase water they don't have: Two-thirds of America's planned AI data centers, 517 of 809, are going up on land that has been in drought all year, and each can burn millions of gallons a day.
An AI designed a vaccine and it's already in people: Cambridge built the first vaccine ingredient made entirely by AI, aimed at the whole coronavirus family, now testing in 39 volunteers.
Your staff is what's blocking AI payoff: A survey of 150 executives put employee fear and distrust as the top barrier to scaling AI, with nearly 60% now hiring around skills.
Anthropic's most dangerous model could ship tomorrow: Leaks and prediction markets point to a public release of its exploit-finding Mythos model as soon as this week, though Anthropic won't confirm a date.
Learn more about AI from the experts building it
📸 Follow us on Instagram for fast, visual AI updates in 30 seconds.
🐝 Subscribe to our Atlas newsletter — trusted by 3M+ subscribers — to stay ahead of AI news across tech, education, and business.
📺 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.
🚀 Explore investment opportunities in the future of AI and join our community-backed growth journey.
