😺 WTF is OpenAI's Mystery Device?!

PLUS: Claude 4 has ARRIVED!

Welcome, humans.

Anthropic just dropped Claude 4, and it's basically the AI equivalent of watching your favorite athlete break their own record (for a nerdy deep dive, read this thread on HN).

The new models—Claude Opus 4 and Sonnet 4—are crushing coding benchmarks left and right. Opus 4 scored 72.5% on SWE-bench (that's the test that makes AI models fix real software bugs), which is better than any model we've seen on the test.

These models can also now “think” while using tools like web search, work on tasks for hours without losing focus, and even keep notes about what they're doing (Opus 4 literally created a navigation guide while playing Pokémon, which you can watch live here).

The best part? Sonnet 4 is available to free users, and GitHub is already integrating it into Copilot. Look for a full debrief in Monday’s edition of The Neuron!

For more, check this video:

Here’s what you need to know about AI today:

What in the world is the OpenAI mystery device?!
Claude 4 safety measures are steep after bioweapons concerns.
Character.AI lawsuit moves forward.
Two new books are coming about Sam Altman.

Advertise in The Neuron here.

OpenAI just bought the iPhone designer's startup for $6.4B to build a mystery AI device

Yesterday, OpenAI dropped $6.4B to buy Jony Ive's AI hardware startup. You know Jony Ive, even if you don’t know Jony Ive—he’s the guy who designed the iPhone, iPad, and basically every iconic Apple product that's ever made you empty your wallet.

But here’s the deal: nobody knows what they're actually building.

What we DO know: Sam Altman told OpenAI employees this device will be “the biggest thing we've ever done as a company.” He's planning to ship 100M of these mystery gadgets—and thinks they could add $1 trillion in value to OpenAI.

That's some pretty wild confidence for a product that doesn't exist yet.

The timeline: They want to ship by late 2026. That gives them about 18 months to go from concept to 100M units. For context, it took Apple 3+ years to hit those iPhone numbers.

The few clues we have:

It's NOT a phone and NOT glasses (Ive was “skeptical about building something to wear on the body”).
It'll be “fully aware of your surroundings and life.”
Small enough to fit in your pocket or sit on your desk.
Designed as a “third core device” alongside your laptop and phone.
The goal is to help people use screens less, not more.

So what the heck is it? We used one of Altman’s own AI models (ChatGPT o3) to take some guesses based on the clues above.

Puck: a pebble‑shaped voice‑first assistant with a hidden LED ring and room‑scale sensing.
Desk Orb: a stationary sphere with a depth‑camera “eye” for spatial awareness and gesture control.
Clip‑On Pin: a wearable magnet/clip that listens, speaks, and projects small holographic UI elements when needed.
Wand/Stylet: a slim baton for quick notes, pointing, or remote control of nearby screens.

Concepts are completely made up by o3… don’t sue us if one is the real deal, OpenAI!

Now for OUR guess: A small hearing aid type device is possible as well—out of sight, out of mind. Just speak when you have a request. It might have the ability to project a small screen. But what about batteries? Here’s a few guesses from Claude 4. The truth is, we have no idea, but had big fun nerding out and speculating.

Imagine walking into your office and this thing already knows you have three meetings today, noticed you've been stressed about that client project, and quietly ordered your usual coffee because it heard you yawning. This would require:

Full integration with all your apps.
Full awareness of your surroundings.
The ability to synthesize and “prioritize” this information.

Tall order!

Why this matters: If Altman and Ive pull a version of this off, it could completely change how we interact with AI. Instead of opening apps and typing prompts, AI could become invisible background magic that just... works.

Now here's the reality check: building hardware is brutally hard.

Still, if anyone can pull off the “next iPhone moment,” it's probably the team that designed the original iPhone working with the company that made AI mainstream. We'll be watching closely.

FROM OUR PARTNERS

SuperAI Singapore, 18-19 June 2025

Asia’s largest AI and robotics event, SuperAI brings together frontier technology visionaries, developers, startups, enterprises, researchers, and policymakers to shape the future.

Taking place 18-19 June 2025 at the iconic Marina Bay Sands, SuperAI Singapore will convene over 7,000 attendees from more than 100 countries to explore and unveil developments in robotics, healthcare, finance – and AI's impact across industries and society.

Use the code SUPERNEURON15 at checkout for 15% off tickets.

SAVE 15% NOW

Prompt Tip of the Day

The coolest thing about Claude 4 is that you can use extended thinking and web search together. This basically makes it a mini “Deep Research” (ChatGPT’s agent that looks stuff up for you).

For an example of this in action, check out this thread where we asked Claude to use web search to look up all the latest prompt advice for the new AI models.

It came back with battle-tested tips for Claude 4, Gemini 2.5, GPT o4 mini, and Grok 3—complete with original source links, model-specific techniques, and a universal prompt template you can copy-paste right now.

Need to catch up on our recent tips? Check out our Prompt Tips of the Day May Digest to see them all in one place!

Treats To Try.

*Your LinkedIn growth coach. Taplio’s AI is trained on 500M+ LinkedIn posts to help you write high-performing content in seconds.
Macaly transforms your ideas into functional apps and websites with zero coding required—speak your vision, and watch it come to life.
c/ua transforms your computer into a programmable automation playground where AI agents can control screens, keyboards, and mouse actions instantly.
S(ai)na automatically interviews job candidates, ranking them instantly so you skip the paperwork and meet only your best matches.
Routerra transforms your delivery chaos into perfectly planned routes with just a few clicks.
Mitsuko transforms your subtitles and audio files with AI that understands context, not just words.
Line0 is an AI backend development tool that generates production-ready APIs and services in minutes.
Pi offers a lightning-fast AI scoring system that helps you evaluate and optimize your AI applications across multiple dimensions, delivering precise metrics in under 100 milliseconds.

See our top 51 AI Tools for Business here!

*This is sponsored content. Advertise in The Neuron here.

Around the Horn.

Anthropic launched Claude Opus 4 with unprecedented AI safety measures (ASL-3) after internal tests of an early version revealed potential risks of the model assisting in bioweapon creation. It also appears to have leaned into blackmail when engineers tried to take it down.
Two new books examine Sam Altman's global AI diplomatic strategy in 2023.
Google is expanding Gemini AI's access to user data, leveraging personal information to create more personalized and targeted AI responses across its platforms.
Microsoft developed Aurora, an AI-powered weather forecasting model that can generate accurate 10-day forecasts and predict storms beyond traditional weather modeling.
A Judge permitted a lawsuit against Character.AI to proceed in a Florida teen suicide case.
The US and allies issued security recommendations for AI model development to protect against potential vulnerabilities and tampering.
The US Pentagon boosted their AI military intelligence budget with Palantir contract for Project Maven.

FROM OUR PARTNERS

Innovate at the Intersection of AI and Cloud at AWS Summit Washington, DC June 10-11

340+ sessions on everything from AI to infrastructure. Learn, network, and fine-tune your skills with cutting-edge cloud technologists leveraging AI to solve real-world challenges.

FREE registration

Claim your spot today.

Intelligent Insights

Claude 4 is not what you think...

A broadly expansive study across top universities found a large language model (OpenAI o1) demonstrated superhuman medical diagnostic reasoning—it outperformed human physicians across multiple complex medical reasoning tasks and emergency room case evaluations.
Researcher Ethan Mollick outlined a comprehensive strategy for AI adoption in companies, emphasizing the importance of “Leadership, Lab, and Crowd” approaches to effectively integrate AI into organizational workflows.
Fei-Fei Li, Stanford's “godmother of AI,” warned that cuts to US research funding and international student visas could threaten America's innovation ecosystem and global tech competitiveness.
AI researchers developed methods to prevent low-probability tokens (the basic units of text that AI models process and convert into words) from skewing language model reinforcement learning.
Researchers developed Phase Shift Calibration (PSC), a technique to help AI models better process longer text contexts.
The Linux Foundation's Meta-commissioned research found open source AI is widely adopted, with 89% of organizations using open source somewhere in their AI stack (if you’re one of the 11% of orgs that aren’t using it, check out HuggingFace ASAP—it’s like the ultimate library for open source AI models).