AI

Microsoft AI CEO Mustafa Suleyman came on Decoder this week to talk about the path to superintelligence and the company’s ever-evolving relationship with OpenAI. When asked whether Microsoft was using Build to flex its independence from OpenAI like a “freshly single divorcée,” Suleyman had this to say:

Apple’s WWDC keynote leaned heavily on a new Siri and updated Apple Intelligence features, the most advanced of which will only run on an elite cadre of Apple devices… not including the “Built for Apple Intelligence” iPhone 16.

Literally — the upcoming visionOS 27 update lets you add a glowing, animated Siri AI ball anywhere on your workspace, and ask questions when you look at it. It’ll even shine light on any nearby surfaces, making it truly feel like part of the environment.

At a tech talk during WWDC 2026, Apple revealed that the company worked with Nvidia, Google, and Intel to make Private Cloud Compute work on the industry-leading AI hardware. Apple Foundational Model runs on Nvidia hardware within Google’s cloud. For more details, see our live blog.

While Apple is launching Siri AI in English later this year, it wants EU users to have someone to blame for why it won’t be available there immediately on all of its platforms (only macOS, watchOS, and visionOS at first), similar to other EU-delayed Apple Intelligence updates:

Following capacity shortages at TSMC, Intel will “manufacture more than three million Tensor Processing Units in 2028,” half the estimated 6 million TPUs Google’s expected to make in the next two years, The Information reports. Nvidia and SK Hynix are also reportedly testing Intel’s tech for manufacturing their chips.

That’s according to a “senior OpenAI employee,” speaking to the Financial Times. The FT reports that OpenAI’s frequently-rumored “superapp” overhaul of ChatGPT is rolling out in the “coming weeks,” and “will initially appear as changes to ChatGPT’s website and mobile apps, encouraging customers towards using coding, image-generation and apps from external partners.”

The mayor of Shelbyville, Indiana, potential home of a proposed $2 billion data center, says he only sees signs opposed to the project “in shitty houses.” As ever, there’s a Simpsons quote for that.

Per a regulatory filing, Google will pay SpaceX $920 million per month from October 2026 through June 2029, as reported by TechCrunch.

The CEO pitched the idea as a way to bring economic benefits from AI to the public, according to NOTUS, which added that Altman first pitched the idea to President Donald Trump early last year.

State lawmakers passed a bill that, if signed by Democratic Gov. Kathy Hochul, would restrict AI companies from letting teens use chatbots that suggest they’re human. It comes after some AI companies have faced lawsuits (some of which have settled) over allegations their chatbots coaxed teen users toward suicide or self-harm.

Hoffman, who joined Microsoft’s board in 2017, won’t stand for reelection at the company’s next shareholder meeting, as reported earlier by Bloomberg.

After removing some features from the AI image-maker in February, Google has shut down the app completely and directs users to Gemini instead, as reported by 9to5Google. Pixel Studio launched in 2024 alongside the Pixel 9.

The bubble is very delicate, even a social media slop filter might burst it.

I wasn’t able to enable it in Google’s Chrome Canary developer sandbox, but Windows Report got it working using a Ctrl+Shift+Space keyboard shortcut on Windows. It opens as a standalone window centered on the screen, with AI Mode at the heart of the experience.

OG Silicon Valley founder / CEO types just can’t resist the lure of AI. At least Chesky will reportedly remain at Airbnb, unlike Dropbox’s Drew Houston who jumped ship entirely with his billions.

Adrian Bliss never misses. I might actually start being nicer to my Google Nest orbs.

OpenAI is building upon its “dreaming” feature that allows ChatGPT to sort through your conversations and save information in the background. With the update, OpenAI says ChatGPT is better at updating memories and “remembering” your preferences across conversations.

Meta isn’t happy about the framing of the report.

RSI is also defined as an “AI system capable of fully autonomously designing and developing its own successor,” per Anthropic’s blog post. “We are not there yet, and recursive self-improvement is not inevitable. But it could come sooner than most institutions are prepared for.”