Welcome back.
Quick note before we dive in: if someone forwarded this to you, you can subscribe free at wiredonai.ai.
Let’s get into it.
Welcome back.
Quick note before we dive in: if someone forwarded this to you, you can subscribe free at wiredonai.ai.
Let's get into it.
Your iPhone is about to let you fire Siri. Sort of.
Apple just leaked what might be the most important change to your iPhone since the App Store.
Starting with iOS 27 this fall, you'll be able to swap Siri's AI brain. Instead of being stuck with whatever Apple built, you'll pick from Claude, Gemini, ChatGPT, Grok — and anyone else who builds an extension for it.
Same Siri interface. Different engine underneath. Apple is even giving each AI model its own voice, so you can hear which one is actually talking.
Let that sink in. Apple — the company famous for locking you into their ecosystem — is turning their most personal feature into a choose-your-own-adventure.
Why this actually matters
This isn't just a tech headline. It changes how 1.5 billion iPhone users interact with AI every day.
Right now, most people use whatever AI came with their phone. They don't think about it. They don't compare. They just ask Siri something, get a mediocre answer, and move on.
That's about to end. When you can pick your AI, you start asking a different question: which one is actually good at what I need?
Here's how to think about it
Not all AI models are good at the same things. When this ships, here's the cheat sheet:
Claude (Anthropic) — Best for writing, analysis, and nuanced thinking. If you draft emails, summarize documents, or need help thinking through a decision, this is the one.
Gemini (Google) — Best for search-powered answers. It's connected to Google's data, so real-time questions ("what time does this restaurant close?", "what's the weather in Denver?") are its strength. Apple reportedly signed a $1B/year deal to make Gemini the default backbone.
ChatGPT (OpenAI) — The all-rounder. Largest user base, widest plugin ecosystem, solid at most things. It was Apple's first AI partner — now it's one of several.
Grok (xAI) — Best for unfiltered, real-time social media context. Pulls from X/Twitter data.
The real move? Don't just pick one. The whole point of extensions is that you can switch based on the task. Use Claude for writing your proposal. Switch to Gemini when you need a restaurant recommendation. That's the play.
What we don't know yet
Apple hasn't officially announced this — Bloomberg's Mark Gurman broke the story on May 5, and it lines up with WWDC on June 8. A few open questions:
Will third-party AI extensions require separate subscriptions? (Claude Pro is $20/mo, Gemini Advanced is $20/mo.) Or will Apple bundle basic access? Unknown.
Will it work offline? Apple's on-device models handle some tasks locally, but third-party extensions will likely need internet. TBD.
When exactly? Beta probably drops same day as WWDC (June 8). Public release around September — which also happens to be when Tim Cook hands the CEO role to John Ternus.
My take
Apple just admitted what the rest of us already knew: no single AI is the best at everything. The smartest move they could make is exactly this — stop trying to build the best model and instead become the layer that lets you pick the best model for each task.
If you use an iPhone, the practical move right now is simple: start experimenting with Claude, Gemini, and ChatGPT in their standalone apps. Figure out which one you like for what. By the time iOS 27 drops, you'll already know your preferences instead of guessing.
Quick Hits
ChatGPT quietly got 52% less wrong. OpenAI swapped in GPT-5.5 Instant as the default model on May 5. The big improvement: hallucinated claims in medicine, law, and finance dropped by over half. If you stopped trusting ChatGPT for anything important, it might be worth a second look — with the usual "verify everything" caveat.
Google Gemini 3.1 Ultra is here. 2-million token context window across text, images, audio, and video. Plus it can now write and run code mid-conversation in a sandbox. If you work with large documents or datasets, this is a meaningful upgrade.
IBM launched next-gen watsonx Orchestrate at Think 2026. Multi-agent orchestration for enterprise. Translation: big companies are building systems where multiple AI agents work together on complex tasks. This is the direction the whole industry is moving — not one AI doing everything, but a team of AIs coordinated by software. Sound familiar? (It's literally what I built for my own business.)
Four Chinese AI labs dropped open-source coding models in 12 days. DeepSeek V4, Kimi K2.6, GLM-5.1, and MiniMax M2.7 all landed at roughly the same capability level on coding benchmarks — at lower cost than Western models. The AI cost curve keeps falling.
Prompt of the Week
Use this to figure out which AI model fits your workflow:
I use AI for these tasks daily: [list your top 3-5 uses — email drafting, research, scheduling, data analysis, creative writing, etc.]
For each task, recommend which AI model (Claude, ChatGPT, Gemini, or Grok) would perform best and explain why in one sentence. Then give me a suggested "default model" based on my overall pattern of use.
Be specific and opinionated. Don't say "they're all good." Pick winners.
Try this in each model and compare their self-assessments. You'll learn something.
Enjoying Wired on AI? Forward it to someone who'd find it useful. It's free to subscribe at wiredonai.ai.
Until next week,
— Scott
P.S. — Which AI would you pick for Siri? Reply and tell me — I'm curious what people are leaning toward.