Sam Altman says OpenAI's first AI hardware prototype is "jaw dropping good" but won't tell anyone what it actually is.

Anthropic quietly dropped Claude Opus 4.5 this week, which scored higher than any human candidate on their toughest internal engineering assessment.

Awkward.

Meanwhile? The federal government just launched a supercomputing mission that could cut research timelines from years to literal hours.

NVIDIA got so rattled by Google's TPU momentum that they put out a whole Twitter post basically saying "we're still the best."

And ChatGPT ads might be coming way sooner than anyone expected.

While you were stuffing your face for Thanksgiving, the AI world continued to cook. 

Don’t miss any important updates. Here's what matters.

According to CNBC, Sam Altman is finally spilling the tea on those mysterious hardware prototypes.

He claims OpenAI’s AI hardware is "jaw dropping good."

Reports say this is the first major move since OpenAI teamed up with legendary designer Jony Ive back in May.

Altman wouldn't say exactly what the device is. But he did paint a picture.

He wants a "calmer experience" than smartphones. Think less "walking through noisy Times Square" and more "sitting in a cabin by a lake."

Sounds nice, right?

But here is the wild part.

Altman suggested the device could filter out distractions because it potentially knows everything you have ever thought, read, or said.

Yuuuuup.

Everything you have ever thought.

The device is likely two years away, but it is clearly aiming to replace your iPhone addiction with something much smarter.

What it means: 

OpenAI is done just being software. They want to be the interface for your entire life.

If a device actually knows everything you’ve thought or said, privacy isn't just an issue. It is the entire product.

Apple should be very, very worried about its former design chief leading this charge.

The White House says the federal government is officially throwing its hat in the AI ring.

President Donald Trump signed a new executive order creating the "Genesis Mission."

This isn't just a small project.

It directs the Department of Energy and its 17 national labs to build a massive integrated AI platform. The goal is to connect current and future supercomputers to speed up scientific discovery.

We are talking about cutting research timelines in health and energy from years down to days. Or even hours.

The administration believes this could change treatment options for fatal diseases within the next five years.

Private companies like NVIDIA and Dell have already shown interest in contributing capacity.

It seems the line between government labs and big tech is about to get very blurry.

What it means: 

Sovereign AI is the biggest trend you aren't watching for 2026.

Governments are starting to view AI compute like oil or gold. It is a natural resource now.

The US is trying to centralize its power to ensure it doesn't fall behind in the global arms race for intelligence.

Did NVIDIA just get defensive?

NVIDIA posted on social media this past week in a rare move to clap back at competitors.

This happened right after their share price dropped 3% following reports that Meta might use Google’s TPUs instead of NVIDIA's chips.

In a post on X, NVIDIA said it was "delighted" by Google's success. But they quickly followed up by reminding everyone that their chips are still the best.

They argued that their GPUs are the only platform that can run all major AI models everywhere.

They also claimed to still control 90% of the AI chip market.

It was a strange PR move. It felt like the industry leader was looking over its shoulder for the first time.

Google's Gemini 3 success and their custom chips are clearly making NVIDIA a little nervous.

What it means: 

The hardware monopoly might finally be cracking.

Big tech companies are tired of paying the "NVIDIA tax" and are building their own infrastructure.

When a company worth trillions feels the need to tweet a defense, you know the competition is getting real.

OpenAI announced this past week that they are rolling out a new shopping research feature.

It is available to everyone. Even free users.

If you ask about a product, ChatGPT now offers a structured, personal shopper experience. It asks follow-up questions about your budget and needs.

OpenAI says this is powered by a version of GPT-5.1-mini tailored specifically for shopping.

But holllllld up.

In our testing, the results were actually pretty bad.

When we compared it to the normal "Deep Research" mode or GPT-5.1-thinking, the shopping bot gave worse recommendations. It even tried to sell irrelevant items like prompt packs instead of software deals.

OpenAI plans to add instant checkout soon, effectively turning the chat into a store.

But right now? It feels half-baked if you’re a paid power user. 

What it means: 

OpenAI is looking for new revenue streams beyond subscriptions.

Affiliate commissions and direct sales are the obvious next step.

But if the specialized "shopping model" is not as good as the general model, users probably aren't going to trust it with their credit card en masses. 

Well, maybe free users will?

Amazon announced this past week one of the largest public sector cloud commitments in history.

They are investing up to $50 billion to expand AI infrastructure for US government customers.

Starting in 2026, the project will add nearly 1.3 gigawatts of power capacity.

Let's be honest. You probably don't know what a gigawatt is. 

That is enough power to support nearly one million homes. All focused on government AI workloads.

This expansion covers AWS Top Secret and AWS Secret cloud regions. It gives federal agencies access to tools like SageMaker and models like Claude.

AWS CEO Matt Garman says this aims to remove tech barriers that have held the government back.

It creates a direct pipeline for the military and intelligence agencies to use the world's best AI.

What it means: 

The scale of energy required for AI is becoming unfathomable.

Powering government AI will require the energy equivalent of a major city.

This confirms that the bottleneck for the next decade won't be chips or code. It will be having enough electricity to turn the machines on.

Anthropic released its most powerful model to date just a week after everyone else dropped theirs.

Claude Opus 4.5 is here.

Anthropic says this model scored higher than any human candidate they have ever interviewed on their internal engineering assessment.

It achieved an 80.9% accuracy on the verified SWE-bench. That beats OpenAI’s GPT-5.1 Codex Max and Google’s Gemini 3 Pro.

It is also waaaaaaay cheaper than their previous pricing approach on the API side. 

The pricing is $5 per million input tokens and $25 per million output tokens. That is a dramatic price cut for a frontier model.

Early testers say it has better intuition for complex workflows and uses 76% fewer tokens to get the job done.

What it means: 

The "smartest model" crown continues to change heads every single week.

For developers and finance pros, this is the new gold standard.

But for everyone else? It proves that intelligence is becoming a commodity. Prices are crashing while capabilities are skyrocketing.

We gave Opus 4.5 the hands-on treatment last week and weren’t super impressed.

Reports just came out that ads are hidden in the ChatGPT code.

So much for a free lunch shorties. 

An engineer named Tibor Blaho found code references in the Android beta version that explicitly mention "sponsored content" and "ads features."

This signals a massive shift for OpenAI.

Why now?

Well, Forbes reports the company needs an additional $207 billion to fund its growth. They aren't expected to be profitable until 2030.

That is a lot of burned cash.

With 800 million weekly active users, ChatGPT is one of the only major platforms not running ads yet.

Sam Altman even hinted on a podcast recently that he finds Instagram ads effective.

It looks like the era of ad-free AI is coming to an end.

What it means: Your "therapist" is about to start selling you stuff.

People tell ChatGPT their deepest secrets. That data is more valuable than anything Google or Facebook has.

Monetizing that intimacy is a financial no-brainer, but it might destroy user trust overnight.

We recently tackled the ChatGPT ads issue in depth, so make sure to go check that episode here.