Welcome, humans.

Apple’s AI era starts now?

We’re going LIVE right now, now that Apple has wrapped its biggest event of the year, and this one is bigger than yet another iPhone launch.

It’s John Ternus’s first major keynote as Apple CEO, and the question we care about is simple: does Apple finally look like an AI company?

We’ll break down what Apple actually showed around Siri and Apple Intelligence, what feels different under Ternus, and whether Apple is finally closing the gap with the rest of the AI industry (lol… ).

The AI bits we’re digging into:

• Siri AI is not fully “done” yet: the English beta starts with iOS 27 on September 14, French, Japanese, Korean, Portuguese, and Spanish follow in October, and Apple hasn’t given a non-beta release date.

• iPhone 18 Pro and Pro Max go up for preorder September 12 and become available September 18. A20 Pro moves to 2nm and 32 neural-processing cores, pushing more advanced AI work directly onto the phone.

• iPhone Duo arrives later: preorders open October 16, with availability starting October 23. Apple also said it used AI in manufacturing to tune the hinge components so they work together precisely.

• Apple is pairing AI creation with provenance: iPhone 18 Pro supports SynthID plus a new Apple Reference Image system designed to help verify what the camera actually captured.

• Some server-heavy Siri features will have usage limits, with increased access through eligible iCloud+ plans.

In other words: Apple’s AI pitch is finally moving from keynote promise to something we can actually use.

Then the rest of today’s AI news gets weird fast

• Meta Muse: Meta launched a personal AI agent with a massive free usage tier, potentially up to 100M tokens per week.

• OpenAI’s $1M math fight: OpenAI says a huge AI effort made progress on Navier-Stokes, one of math’s Millennium Prize problems, and immediately sparked a fight over credit.

• AI risk: An Anthropic scientist put the chance of AI causing human extinction above 10%.

• Microsoft’s monster patch: Microsoft fixed a record 972 vulnerabilities, including 112 rated critical.

• U.S. agencies vs. Chinese AI labs: the NSA, FBI, and CISA warned that companies including DeepSeek, Moonshot, and Alibaba are distilling Western AI systems “at an industrial scale.”

And yes, Sam Altman is apparently very excited about an Ocarina of Time remake. Important work continues.

Catch up on Apple’s event here, then come argue with us in the live chat.

Want the full AI rundown? 

We went through Apple’s post-event releases and pulled the models, model sizes and evaluations, Siri AI, A20 Pro, new iPhone Duo manufacturing deets, Apple Watch Audio Intelligence features, and cool health AI add-ons into one place:

Everything AI Apple Announced at Its September 9 Event.

Tomorrow: GitHub for total beginners, LIVE with Cassidy Williams

AI can build you an app before you’ve learned what a repo or pull request is. Tomorrow at 10 AM PT / 12 PM CT / 1 PM ET, Cassidy Williams from GitHub is taking us through GitHub (like Google Drive or iCloud, but for your code) from zero: repos, branches, pushing and merging, AI coding agents, worktrees, GitHub Actions, key management, and how to safely ship what AI built.

If you wanna build stuff with AI, but you’re a total beginner, this one is for you!

Coming Soon From The Neuron: OpenAI Astra Changes Everything, But Not How You Think

We also have a new pre-recorded episode of The Neuron: AI Explained dropping later today, and this one is based on actually living with Astra for a weekend, not staring at benchmark charts.

Corey burned roughly 2B tokens in three days. Grant had Astra orchestrating subagents on one project for eight to nine hours. Corey dictated a game idea into Codex Remote while barbecuing, then came back to a playable build on his phone.

The bigger thread: as frontier models get more capable, the bottleneck starts moving away from the model itself and into monitoring, your computer’s hardware, and the infrastructure underneath it.

Corey and Grant get into:

• Why Astra’s recurrent-depth reasoning makes hard problems easier and cyber safeguards more consequential.

• Gemini 3.8 Flash, Flash Cyber, and Corey’s new benchmark rule: “my benchmark is my work.”

• Why computer-using agents could make CPU performance matter again, plus Mac Studio vs. DGX Spark.

• Why Texas is slowing the data-center rush, and what communities should demand before saying yes.

• ChatGPT’s reported $1B ad run rate, and what happens if AI answers stay separate from ads.

Also: a $399 open-source robot duck and Dyson’s $500 CameraJet AI toothbrush. Because apparently all roads to AGI pass through consumer electronics.

🎙️ ICYMI!

A few recent episodes and streams worth catching up on:

1. Is OpenAI about to launch GPT-6 / Astra? We broke down what OpenAI had actually confirmed, what looked like rumor, and why Astra’s cyber capabilities matter.

2. Claude Fable 5.1 LIVE We tested Anthropic’s newest agent model on coding, research, writing, and long-running autonomous work.

3. AI agents have a security problem nobody has solved Alice CEO Noam Schwartz explains what changes when agents get tools, credentials, memory, and permission to act.

4. AI Tool Roundup: What’s Actually Worth Trying? A plain-English tour through the models, local AI tools, and agent products that were actually worth your attention.

Stay curious,

The Neuron Team

P.P.S: Love the newsletter, but don’t want these podcast and livestream announcement emails? Don’t unsubscribe. Adjust your preferences to opt out of them here instead.