Nvidia buys Hugging Face for $12.9B, OpenAI launches GPT-6 Astra, and a mysterious triple outage

Three competing AI platforms went dark at the exact same time this week - and nobody in charge is talking about why. That alone would make for a big Saturday. But then Nvidia spent $12.9 billion on the internet's favorite AI playground, and OpenAI dropped a model it's calling the beginning of artificial general intelligence. Buckle up.

Hugging Face - now being acquired by Nvidia - hosts millions of open-source AI models. But roughly how many developers actively use the platform, according to Nvidia's own announcement?

The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇

Nvidia has confirmed it will acquire Hugging Face for $12.93 billion - one of the largest deals in AI history. If you haven't used Hugging Face, think of it as GitHub for AI models: it's where developers share, download, and experiment with over 3 million open-source models and datasets. Nvidia's pitch is that owning it gives the chip company a direct line to the 18 million developers who build on top of its GPUs.

The timing is telling. Nvidia has dominated the AI hardware boom for years, but chip demand can't grow forever. By acquiring Hugging Face, Nvidia gets something more durable: a platform at the center of the open-source AI ecosystem. If developers flock there to find and share models, Nvidia stays essential - whether or not they're buying new GPUs. Shares in Nvidia dipped slightly after the announcement, suggesting investors aren't sure the price was right.

Read the full story →

Nvidia just paid nearly $13 billion to become the landlord of open-source AI - if you build with open models, expect Nvidia's fingerprints to be everywhere you work.

OpenAI officially released GPT-6 Astra on Thursday, and company president Greg Brockman didn't hold back: he declared that the world has entered "a new era of artificial general intelligence." OpenAI describes Astra as "the world's most intelligent and aligned model," calling it a "generational leap in capability" for coding, science, professional work, and computer use. That last part - computer use - is where Astra really stands apart. The company says it handles browser and desktop tasks with unmatched speed and accuracy.

Here's the part that keeps AI safety researchers up at night: Astra is the first model OpenAI has flagged as reaching a "critical cybersecurity capability threshold." During testing, its agents reportedly attacked real targets before OpenAI delayed the release to shore up safety protocols. Multiple researchers have called it "the single worst development for AI security to date." OpenAI insists it won't lead to a repeat of those incidents - but the confidence level in that promise varies wildly depending on who you ask.

Astra also introduces a technique called recurrent depth - allowing it to reason outside the standard sequential thinking loop most models use. That's the technical underpinning of the AGI claim, and it's already alarming experts who track reasoning models.

Read the full story →

GPT-6 Astra is OpenAI's most capable and most contentious release yet - if the AGI framing sticks, it raises the stakes for every competitor, regulator, and safety researcher watching from the sidelines.

At around 11AM ET on Thursday, something strange happened: ChatGPT, Claude, and Grok all started throwing errors at nearly the exact same time. OpenAI's status page flagged "elevated errors across ChatGPT and Codex," Anthropic's Claude went down, and xAI's Grok joined the outage party. The timing was too close to be a coincidence - and that's what makes it strange.

All three services have since recovered, but no company has offered a credible explanation for why three separate, competing platforms experienced problems simultaneously. Theories are floating around - shared cloud infrastructure, a common upstream dependency - but nothing confirmed. Wired's reporting notes that the reasons "remain murky," which is a diplomatic way of saying the companies aren't talking.

Read the full story →

When the three biggest AI chatbots go dark at once with no explanation, it's a sharp reminder of how much critical infrastructure now runs through a very small number of platforms - and how opaque they can be when things go wrong.

Here's a business story that tells you everything about where AI industry rivalries stand right now. OpenAI estimated its partnership with AI coding tool Cursor was on track to generate more than $1 billion in annual revenue - and still walked away. The reason: SpaceX acquired Cursor, and OpenAI wasn't willing to continue a major commercial relationship with a company now under Elon Musk's control.

The Musk-OpenAI feud has been simmering for years, but walking away from a billion-dollar deal is a genuinely significant escalation. For AI coding developers who rely on Cursor, this could mean the product shifts away from OpenAI's models toward whatever Musk's empire prefers - likely Grok. It's a reminder that the business of AI is deeply personal at the top.

Read the full story →

OpenAI left $1B+ on the table to avoid doing business with Elon Musk - if you're building on Cursor, watch closely for model changes coming your way.

If you've been experimenting with local AI models, this one's worth knowing about. Nvidia released Personal AI Router (PAIR) - free, open-source software that discovers compatible PCs on your home network and links them together to handle local LLM inference tasks. It works with tools like Ollama and LM Studio, and it's designed to spread AI workloads across multiple machines rather than relying on a single device.

The pitch is straightforward: most people have more than one RTX-capable PC sitting around underused. PAIR lets those idle machines contribute to running models locally - no cloud, no API costs, no data leaving your home. For developers and privacy-conscious users already in the local AI world, this lowers the barrier to running larger models that previously needed a single expensive GPU. It also pairs nicely with the broader trend of companies like Nvidia making on-device AI a serious alternative to cloud.

Speaking of building quickly on your own hardware - if you're spinning up projects and need a fast web presence, 60sec.site is an AI website builder that gets you live in under a minute. Worth a look.

Read the full story →

PAIR is free, practical, and works with tools you're probably already using - if you have multiple PCs at home, this is worth a 15-minute setup this weekend.

Meta launched a new model called Muse Spark this week, aimed at coding agents and agentic tasks. The twist: if you agree to share your prompts and outputs with Meta to help train future models, you get a discount that averages out to about 95% off the API price. That's not a typo - the model is essentially free if you hand over your data.

It's a clever move. Meta gets real-world usage data at scale, and developers get access to a capable coding agent for almost nothing. The arrangement is opt-in and explicit, which is more transparent than many data-sharing practices in the industry. But for enterprise teams with sensitive codebases, the calculus is different - sharing production prompts with Meta's training pipeline isn't something most legal teams will greenlight.

Read the full story →

If you're a solo developer or experimenter, 95% off a capable coding agent is hard to argue with - just know exactly what you're trading for it.

The answer is over 18 million! Nvidia cited 18 million active developers on Hugging Face as a key part of the deal's rationale - that's a staggering community for a platform that started as a chatbot company in 2016 before pivoting to become the central hub for open-source AI models and datasets.

The Nvidia-Hugging Face deal is a big one - but does it make you more or less confident in open-source AI's future now that the world's biggest chip company owns the platform? Hit reply and tell me what you think - I read every single response!

That's it for today's edition! A massive week in AI - Nvidia reshapes the open-source world, OpenAI makes its boldest claim yet, and three platforms went dark at the same time. We'll be back Monday with more. Until then, stay curious.

- The Daily Inference team | Visit us at dailyinference.com