OpenAI recently hired two dozen Apple hardware veterans to build multiple AI devices while a recent high school graduate reportedly now runs XAI's data annotation team.

Oh, and Microsoft's CEO straight up told employees AI might make their entire company irrelevant. Yeah, that Microsoft.

Strange AI timeline we’re living in.

That kind of week in AI, y'all.

Here's what matters, y’all. 

How badly did Apple screw up their AI strategy?

So bad that OpenAI hired more than 24 Apple veterans since early this year. Reports from The Information show this isn't about AI researchers.

Nope.

Hardware people. Interface design, wearables, cameras, audio, manufacturing design, supply chain.

This recruiting wave follows OpenAI's May 2025 acquisition of io Products. You know, the hardware startup co-founded by former Apple design chief Jony Ive and Tang Tan.

Tan's now OpenAI's chief hardware officer.

Notable recruits include Cyrus Daniel Irani who spent 15 years at Apple. Matt Theobald with 17 years. Eric Dijon from the Apple Watch hardware team.

And OpenAI isn't just building one device.

They're exploring a screenless smart speaker, smart glasses, a wearable pin, and a digital voice recorder. Basically trying to figure out the perfect always-available AI form factor.

But here's where it gets spicy, y'all.

Luxshare, a major iPhone assembler, signed on to handle assembly. Goertek got tapped for components.

OpenAI is literally using Apple's own suppliers against them.

Apple got so spooked they canceled their annual China meeting in August to reduce poaching risk.

What it means: Apple has multiple class action lawsuits for promoting AI it cannot deliver.

Now they're losing senior hardware leaders too.

Almost everyone seems to be jumping ship from Cupertino.

Wait, the US government picked Meta over Musk?

Yep.

The US General Services Administration just added Meta's Llama AI to their approved tools list for federal agencies. Reuters reports this clears the way for government-wide experimentation and deployment.

GSA already approved Amazon Web Services, Microsoft, Google, Anthropic, and OpenAI.

But not xAI's Grok.

Awkward.

Especially considering the whole Trump-Musk bromance that turned into a very public breakup. Federal agencies can now test Llama for free under GSA's assurance that it meets government security and legal standards.

What it means: 

The government is literally using every single AI model except Grok after the very public Trump-Musk breakup.

Federal agencies now have access to Meta's multimodal capabilities.

This creates massive validation for Meta's enterprise AI credibility.

Google's rolling out Gemini-powered features directly inside Chrome on desktop right now in the US.

No downloads. No subscriptions. No extra hoops.

There's a new button in the top right of Chrome that launches the Gemini chatbot.

You can unpin it if you hate it.

The killer feature? Cross-tab synthesis.

Gemini can answer questions about the page you're on and pull context from across multiple tabs. That can cut research time for students, knowledge workers, and small business teams juggling too many sources.

Google's also adding their AI mode called the Omnibox before month's end.

Button and keyboard shortcuts that trigger Gemini suggestions based on your current page.

What it means: 

Google just put AI in front of tens of millions of people who weren't actively seeking it out.

So much of getting the most out of generative AI is about taking advantage of the context you currently have.

Check your Chrome browser right now for that new Gemini button.

New usage studies from OpenAI and Anthropic just dropped.

People use ChatGPT mainly to assist in their work while Claude gets handed entire tasks.

Clear split.

OpenAI reports that "asking" conversations where users seek guidance without direct output have outpaced "doing" and "expressing" queries. Many rely on ChatGPT as a thinking partner versus a task finisher.

ChatGPT's most common uses are writing, editing, summarization, and brainstorming.

Strong adoption by students, marketers, knowledge workers, and educators.

Claude? Different story entirely.

Anthropic's study found Claude is used heavily for coding and math-based tasks. Software development and quantitative work account for more than a third of activity.

For the first time in their tracking, automation usage exceeded augmentation.

Companies embed Claude in workflows for document processing, report generation, and bulk coding.

What it means: 

There's a practical split between the two AI startup giants.

ChatGPT thrives in consumer productivity where users iterate alongside it.

Claude's gaining ground in enterprise settings where teams delegate complete tasks for speed and scale.

Google and more than 60 partners announced the Open Source Agents Payment Protocol, or AP2.

First common standard that allows any AI agent to safely complete purchases with clear proof of human user permission.

Right now it's available to developers via a public GitHub repo.

But no consumer facing implementations yet.

Backers include American Express, Coinbase, Etsy, Intuit, MasterCard, PayPal, Salesforce, ServiceNow, UnionPay, WorldPay, and more. This signals broad industry buy-in across payments, commerce, and enterprise software.

The core mechanism uses "mandates."

Tamper-proof cryptographically signed digital contracts using verifiable credentials that prove a user's intent.

What it means: 

This is the first common standard that allows any AI agent to safely complete purchases with clear proof of human user permission.

Right now if you want to roll the dice and give agents your credit card info, you can do that.

But once this new AP2 protocol comes out, that might change everything for AI agents. And make them WAY more useful.

More than $40 billion has been recently invested into the UK by major US tech firms tied to data centers, infrastructure and quantum computing, reports show.

The UK government will speed up data center approvals and ease grid connections.

Cutting costs and timelines for power-hungry AI projects.

OpenAI, NVIDIA, and UK operator Nscale will build Stargate UK. Starting with 8,000 NVIDIA Blackwell GPUs and eventually scaling up to 31,000.

OpenAI will also launch the OpenAI Academy to provide AI training and reskilling for up to 7.5 million UK workers by 2030.

Microsoft's planning to invest more than $30 billion over 4 years.

Including a supercomputer near London with more than 23,000 GPUs.

What it means: 

People want to sit back and say AI scaling has hit a wall or AI is going to go away or the AI bubble is going to burst.

Well, the biggest tech companies who are creating the technology are continuing to invest tens of billions of dollars worldwide on future projects.

If the people making the technology had any hesitation about this technology slowing down, they probably wouldn't be investing all their pennies in building it out bigger.

xAI reportedly put Diego Pasini, a recent high school graduate and University of Pennsylvania student currently on leave, in charge of its data annotation team.

Business Insider broke this wild story.

This happened after more than 600 people were cut from XAI's data annotation unit.

Headcount dropped from about 1,500 to about 900 after multiple rounds of cuts.

Workers told Business Insider that Pasini said on September 15th that there were no further layoffs planned.

Then they cut another 100+ people right after that announcement.

Yikes.

What it means: 

If you're an enterprise company, how would you feel if someone that graduated high school in 2023 was leading data labeling?

When you look at who the other big players have in similar positions, it's usually people with multiple PhDs and decades of experience in machine learning.

The data annotation team handles labeling and contextualizing raw data used to train Grok, so those staffing shifts touch a mission-critical pipeline for model quality and safety.

Nvidia is investing $5 billion in Intel and launching a joint effort and launched a joint effort to develop chips for PCs and data centers.

Rare collab that could reshape the competitive landscape in AI hardware.

Intel shares jumped 23% on the news.

Massive jump that lifted Intel's market value by about $143 billion.

This move follows fresh outside backing for Intel as the US government recently took a 10% stake worth about $9 billion. Japan's SoftBank invested $2 billion.

Intel has struggled in recent years amid leadership turnover, technical missteps, and tough competition from NVIDIA, AMD, and Broadcom.

What it means: 

This could be one of those pieces that turns Intel around.

Intel has struggled and one of its biggest competitors said they're going to invest $5 billion and work together.

The partnership aims to co-develop chips targeting both consumer PCs and cloud data centers.

How scared should we be when the CEO of the second largest company in the world says this?

Satya Nadella bluntly told employees at a recent internal town hall that AI could make Windows, Office, and even Azure less relevant.

That Microsoft itself could become irrelevant if it clings to past success.

He said he's haunted by Digital Equipment Corporation's collapse. Uses DEC as a warning that great companies can all but disappear if they miss a platform shift.

Nadella said AI already writes content for Excel and PowerPoint and could undercut the value of Microsoft's most profitable products if the company doesn't move faster.

He told employees to stop protecting legacy franchises.

Instead build for what the market will demand next.

The message to teams was clear: ship AI-native products and be willing to disrupt Microsoft's own businesses before competitors do it.

What it means: 

There's an important message here from Microsoft's CEO.

If the CEO of the second largest company in the world is saying AI could make us irrelevant, it's a sobering message.

You have to take AI really seriously and stop hanging on to legacy.

You really have to unlearn and start from scratch.