ReutersReuters

Inside OpenAI's vision for its AI devices

Refinitiv4 min read

By Krystal Hu

(Artificial Intelligencer is published every Wednesday. Think your friend or colleague should know about us? Forward this newsletter to them. They can also subscribe here.)

AI is disrupting everything — and now it’s coming for one of the most human arenas: dating.

The shift is so profound that Justin McLeod, founder and CEO of Hinge, is stepping down to build a new AI-focused dating venture called Overtone.

What’s striking is that McLeod — someone who has spent a decade building one of the most popular dating apps — is both optimistic and deeply uneasy about AI’s role in our relationships. He wrote that AI isn’t just mediating our connections anymore. “AI is now attempting to interact directly with us.”

He understands the appeal — constant companionship during a loneliness epidemic — but warns it comes at a cost:

“Real relationships involve risk, vulnerability, effort and reciprocity…We should be very worried when people start choosing artificial intimacy over the real thing.”

Social media already primed us for overstimulation and shortcuts; AI may supercharge both. How will our relationship with AI evolve if we have an always-on device? In this week’s issue, we get an inside look at how OpenAI is building its first AI device, plus fresh data on how soon businesses expect to adopt AI. Scroll on.

Email me or follow me on LinkedIn to share any thoughts.

OUR LATEST REPORTING IN TECH & AI:

  • Exclusive-Nvidia builds location verification tech that could help fight chip smuggling

  • Exclusive-SoftBank, Nvidia looking to invest in Skild AI at $14 billion valuation, sources say

  • How Australia is enforcing world-first teen social media ban

  • SpaceX to pursue 2026 IPO raising above $25 billion, source says

  • Silicon Valley-backed defense firms face growing pains after hot streak

HOW OPENAI’S BUILDING ITS AI DEVICE

OpenAI CEO Sam Altman and legendary Apple designer Jony Ive are building one of the most anticipated AI products: a new consumer hardware device, a collaboration Altman's company has already paid $6.5 billion for.

With a prototype now circulating internally, the whole industry is wondering whether OpenAI can recreate the magic of the original ChatGPT moment — this time in hardware. Based on conversations with people familiar with the device, here’s what we know.

The core vision is simple but radical: AI needs full context. Unlike our phones—which are either on or off, in a pocket or on a table—this device would understand the world continuously, like a truly proactive assistant. OpenAI’s device aims to be always present, always sensing, but with explicit, visible signals that show when it’s paying attention.

Altman’s north star is a device that can listen to your life, understand what you’re doing, handle the majority of follow-up tasks, and surface only what you actually need.

To power these devices, OpenAI’s eventual vision isn't just giant cloud-based AI systems, but small models that can run meaningful AI locally. While OpenAI built its reputation on massive, compute-hungry models, the rapid progress of its compact “Mini” models has reshaped its roadmap.

Insiders say these local models will be critical for a device that is listening and watching, and help address privacy concerns: most people won’t want their entire life streamed to the cloud.

To make that possible, OpenAI will need a new kind of chip. Today’s server chips — Nvidia’s or anyone else’s — are optimized for parallel workloads across millions of users. A personal AI device needs the opposite: a processor built for one user, with tight power constraints and the ability to run compact models in real time. OpenAI plans to work on exactly that: a custom chip optimized for on-device inference.

This family of devices will roll out in phases. The lighter, task-specific and cloud-based devices will come sooner. The more privacy-sensitive, always-on devices will arrive later, as insiders caution the powerful on-device computer might take a few years to mature.

OpenAI is entering a field that's already heating up. Google announced this week that it’s partnering with Warby Parker on lightweight AI-powered glasses launching in 2026. Meta acquired Limitless, the startup building “AI memory” wearables that continuously record and summarize your day. Ambient hardware is becoming a battleground.

Which tech giant pulls this off first, and whether consumers are actually ready for an always-on AI device, will define the next era of AI.

CHART OF THE WEEK:

Thomson ReutersAI adoption still has a long way to go

A reality check from the U.S. Census Bureau shows where we actually are in the AI adoption curve. The survey finds that nearly 57% of businesses have no plans to use AI in the next six months, while another 22% aren’t sure if they will. Only 21% expect to adopt AI tools in the near term. Against the backdrop of the feverish capital pouring into the sector, it’s a reminder that outside Silicon Valley demos and boardrooms, AI adoption in the real economy is still in its early innings.

Investors echoed that caution mixed with optimism at Abu Dhabi Finance Week, where several of the world’s biggest investors warned that AI valuations are running far ahead of fundamentals. Hedge fund manager Chris Hohn called today’s environment “a disruptive force” with rising uncertainty, while Franklin Templeton CEO Jenny Johnson likened the moment to the early days of a gold rush — still years away from delivering real earnings impact.

Login or create a forever free account to read this news