Welcome, humans.

iOS 27 starts landing on iPhones today, and the headline feature is Siri AI: a full app makeover, on-screen awareness, and a language model built with help from Google's Gemini.

We've been waiting for a Siri like this since the iPhone 15 launched. Apple first promised the smarter version for iOS 18 back in 2024, then pulled it because it didn't work reliably enough. Better three years late than never, we guess. Apple's basically the friend who shows up to the party as everyone else is leaving, but at least they brought good snacks this time.

Check Settings, General, Software Update today to grab it. Heads up: the full Siri AI experience needs an iPhone 15 Pro or newer, so a standard iPhone 15 or older gets the update but not the upgrade.

Here’s what happened in AI today:

😾 Nadella, Amodei, Altman, and Musk all agree: slow down AI.
📰 King Charles will host AI chiefs at a slowdown summit.
📰 David Sacks accused Anthropic's whistleblower of a PR stunt.
🍪 Naseem lets an AI agent actually run your Mac.
🎓 Microsoft Copilot has a secret brain swap.

😾 AI's Biggest Rivals Just Agreed on Something: Slow Down

DEEP DIVE: Can AI Companies Afford to Slow Down?

This week, the three men racing hardest to build the world's smartest AI all said the same thing out loud: slow down.

Here's what happened:

Dario Amodei (Anthropic) published an essay warning AI has been "advancing drastically faster" than expected, and called on labs to "pace the frontier."
Sam Altman (OpenAI) and Elon Musk (xAI) both backed him within a day. Musk's entire response: "Dario is right."
Amodei went further on CBS: "I won't lie to you, there are real dangers," and for too long the industry lied about them.
Microsoft CEO Satya Nadella joined in, writing that superintelligence (AI smarter than humans at nearly everything) not kept under human control "is not worth pursuing."
King Charles is hosting Nvidia, Google DeepMind, OpenAI, and Anthropic bosses in Scotland this week to hash out shared safety principles.

Our take: It's genuinely alarming when the people building this stuff say it out loud. Superintelligence is scary because it means AI smarter than any human at nearly everything, not just chess or code. We've spent decades telling ourselves stories about exactly this, from HAL to Skynet, and sure, fiction tends to exaggerate for drama. But strip away the killer robots and the core idea holds up: something this powerful can make things, or break them, and there isn't much room in between.

That's also why we're not popping champagne yet. Three rivals just agreed to slow down together, and it's not purely selfless: Anthropic is reportedly chasing a $2 trillion IPO, and safety staff privately fearing the tech is dangerous creates a real disclosure problem. It's also why David Sacks spent Friday's All-In podcast accusing Jacob Coxon, the researcher who started this conversation, of running a PR stunt, not blowing a genuine whistle. Bernie Sanders wants superintelligence banned outright, and is pushing Trump and Xi to negotiate that at their summit.

And then there’s Trump, who is very much not joining the group hug. He said the U.S. can add “guardrails,” but slowing down risks handing the AI lead to China: “whoever wins AI, wins.” So even if the labs suddenly agree on the danger, the White House is still looking at this like a race you cannot afford to lose. That makes “slow down” a lot harder in practice than it sounds in an essay.

And this conversation didn't exactly start this week. Pope Leo XIV was already warning about the same thing back in May, saying AI needs to stay under human control. So now you have the Pope, Microsoft, Anthropic, OpenAI, xAI, and apparently King Charles all circling the same question: how far do we actually want this to go? The Dune fans among you are probably already thinking “Butlerian Jihad.” Let's maybe not take the reference that literally. What I actually want to see now is whether all this talk changes anything. Do the labs really slow down, change their roadmaps, or put new limits in place? Or does everyone say the responsible thing in public and keep racing behind the scenes?

FROM OUR PARTNERS

If an AI agent made the wrong call yesterday, could you reconstruct what happened and who approved it? Most risk leaders cannot.

The governance infrastructure most organizations rely on was built assuming a human operates the system and stays accountable for it. That assumption no longer holds. Agents are taking action across enterprise systems on their own, and most organizations have no way to trace the decision back.

This guide covers the five sources of enterprise AI risk and what you need in place before your board or regulator asks the question first.

Download the Guide

🎓 AI Skill of the Day: Microsoft Copilot Has a Secret Brain Swap

Buried inside Microsoft Copilot is a toggle almost nobody touches: which AI actually answers you.

Since earlier this year, Microsoft has quietly let Microsoft 365 Copilot users swap out the default OpenAI model for Claude, Anthropic's rival AI, on specific features. The clearest one is Researcher (Copilot's deep-research agent, which digs through your emails, files, chats, and the web to answer complex, multi-step questions). Different models write and reason differently, so if a Copilot answer feels flat, the fix might be a different brain, not a better prompt.

Open Researcher inside Copilot Chat and start a query.
Look for a model picker near the input box, sometimes labeled "Try Claude."
Switch it from the default model to Claude Opus, then rerun the same question.
Don't see the option? Ask IT to enable Anthropic models under Copilot settings in the Microsoft 365 admin center; it's off by default in the EU and UK.

Claude tends to write cleaner long-form analysis; the default model is often faster for quick lookups. Worth testing both before you assume Copilot just "isn't that smart."

Research [TOPIC], pulling from my emails, files, and the web. Compare [OPTION A] to [OPTION B] and give me a one-page recommendation with sources.

Have a specific skill you want to learn? Request it here.

FROM OUR PARTNERS

Some teams never seem to stop moving. They're on Attio, the agentic CRM.

It’s your always-on revenue engine: agents and workflows build pipeline, chase every buying signal, and move deals forward alongside your team.

Teams like Parallel, Turbopuffer, and Wordsmith build on Attio. Are you one of them?

Try Attio now

🍪 Treats to Try

*Discover the potential of artificial intelligence with our comprehensive cheat sheet. Learn more about the concepts, platforms and applications of AI.
Naseem hands your Mac's terminal, files, and even the iOS Simulator to an AI agent that asks permission before it touches anything (free to start, or $29 one-time for Pro through September 30, down from $49).
Sourclip captures webpages, YouTube videos, PDFs, and your AI chat transcripts straight into NotebookLM, then exports the study guides and audio overviews Google won't let you download natively (free to try, then $24/year for unlimited captures).
OpenClaw runs a genuinely capable AI assistant on your own computer, letting you manage email, book flights, and run your calendar from WhatsApp or Telegram (free, open source, no subscription).
Onset turns your finished pull requests into a polished changelog, then pings your team on Slack the moment it's published (no pricing details).
Keiki lets you build one customer-facing AI agent and deploy it everywhere your customers already are: SMS, WhatsApp, Slack, and email (no pricing details).
Nex builds you an AI agent that scores leads against your own closed-won deals instead of generic checklists, then drafts win-back emails for cold accounts (no pricing details).

📰 Around the Horn

Funny timing: the same week Amodei published his "pace the frontier" essay, researchers at Shanghai Jiao Tong and Theseus Labs quietly mapped the actual five-stage roadmap toward the kind of self-improving AI he's worried about, from models that just execute our instructions (L1) to ones that rewrite the mechanics of their own improvement process (L5). Nothing to see here, just AI publishing its own trajectory notes.

King Charles will host Nvidia, Google, and Anthropic bosses at a Scotland AI safety summit this week.
David Sacks accused ex-Anthropic researcher Jacob Coxon of staging his AI-doom resignation as a PR stunt.
Bernie Sanders urged Trump and Xi to negotiate a treaty pausing AI and banning superintelligence outright.
OpenAI, Anthropic, and Google are discussing a shared body for testing and safety standards on frontier AI.
China's state-backed Global Times called Amodei's slowdown proposal a “Cold War playbook” aimed at holding back China's AI progress.