Everything That Happened in AI Today (Tuesday, November 10, 2026)

Anthropic says its newest Claude model built a working operating system with help from more than 1,000 AI agents. The experiment took eight days.

Welcome to Tuesday's Around the Horn, where Anthropic has given Claude some seriously ambitious assignments.

The company has released two upgraded models for complex technical work, and its internal experiments include designing electronic devices and tackling difficult physics problems. Apparently, fixing a few bugs wasn't enough of a challenge.

Elsewhere, Anthropic reportedly sent bankers details of a massive IPO before retracting them. Fireworks is dealing with the fallout from cyberattacks, while NVIDIA is reportedly restricting how overseas partners rent AI chips to Chinese customers.

Let's get into it.

🆕 NEW From The Neuron

• Why Jensen Huang's Affair Has Become an AI Chip Security Concern examines reports that NVIDIA's CEO was targeted by Chinese intelligence and the company's response to the allegations.
• Anthropic's $3.4 Billion Power Deal Shows How Expensive AI Is Becoming explains why Anthropic reportedly committed billions to gas generators for future data centers.
• Adobe Finds 68% of Creative AI Users Sometimes Settle for Results That Miss Their Vision looks at a frustrating problem for AI creators: getting faster results that still miss what they originally wanted.

Around the Horn: Tuesday, November 10, 2026

Anthropic just released Claude Fable 5.2 and Claude Mythos 5.2, its newest models for difficult technical work. The company says the upgrades can tackle complex coding projects and scientific research problems that previous models struggled to solve.

And Anthropic has some interesting examples to back up those claims. During testing, an engineer asked Fable 5.2 to build an operating system from scratch. The model coordinated more than 1,000 AI agents over eight days and produced a working system that could boot on a laptop, complete with windows, file storage, and a login screen.

Anthropic also reported that the models helped solve two previously open problems in black hole physics and designed three working electronic devices.

The models also generate responses up to 30% faster than their predecessors and cost less to run. Both use the same underlying model, but there's an important distinction in who gets access. Fable 5.2 is available to everyone, with safeguards that block potentially dangerous cybersecurity and biology requests. Mythos 5.2 retains those capabilities for verified organizations conducting approved research.

Anthropic is making a big bet on AI systems that can manage complicated assignments over hours or even days. The operating-system experiment gives a glimpse of that ambition, although it doesn't establish that the model can independently handle every real-world engineering project.

We used to ask chatbots to explain computer operating systems. Anthropic is now asking one to build them.

🏆 TOP 5 NEWS (Around the Horn)

• Anthropic circulated and then retracted details of a potential $69.3 billion IPO, according to Reuters. The company reportedly told bankers that the pricing information was inaccurate and that an updated filing would follow.
• Fireworks AI disclosed multiple cyberattacks that exposed customer information and portions of unfinished AI models. Two customers told Forbes they had moved substantial business away from the company amid concerns about security and communication.
• NVIDIA reportedly tightened overseas AI chip rentals to China, requiring cloud partners to reserve at least 51% of their services for non-Chinese clients. The report also said B300 server racks were selling for up to $2.8 million in China.
• Meta expanded AI age detection globally to identify Instagram users aged 13 to 17 and automatically place them in more restrictive teen accounts, even when they registered with an adult birth date. Facebook and Threads will receive wider use of the technology too.
• Qantas announced plans to cut 850 Australian corporate jobs as it outsources technology work to Accenture, with most affected roles expected to move to India. The airline cited IT modernization and the growing use of AI.

Honorable Mentions

• Disney partnered with Adobe to develop custom Firefly AI models for its characters and entertainment productions, with potential applications across film and theme parks.
• Investors reportedly began selling AI chip and data-center loans at steep discounts, amid concerns about rapidly depreciating GPUs and the financial risks associated with AI infrastructure expansion.
• Scott Wiener, Nancy Pelosi's successor in Congress, wants tougher federal AI regulation after championing safety legislation in California.
• Trump claimed his administration threatened an unnamed AI executive with jail time during a 60 Minutes interview. He also expressed openness to AI safety safeguards while warning against excessive restrictions.

🍪 TOP TREATS TO TRY

• Claude Fable 5.2 and Mythos 5.2: Anthropic's new models handle complex coding and scientific research, with responses up to 30% faster than their predecessors. Fable is generally available for $8 per million input tokens and $40 per million output tokens, while Mythos requires approval through Anthropic's verification programs.
• Muse Ember: Meta introduced its smallest, most efficient Muse model for repetitive AI-agent tasks, including tool use, coding subtasks, and document processing. It's available through the Meta Model API for $0.10 per million input tokens and $0.50 per million output tokens.
• Google TPU 8tx and 8i Max: Google unveiled two specialized AI chip systems. The 8tx protects sensitive information while AI processes it, while the 8i Max combines two chips for memory-intensive tasks. Bloomberg reported that the expanded lineup intensifies Google's competition with NVIDIA. Pricing and availability dates remain undisclosed.
• Apple Intelligence Usage Packs: Apple began selling $49.99 one-time packs that provide additional AI usage after monthly limits are reached. Logic Pro is the first supported app, with other Creator Studio apps expected to follow. Unused purchased usage expires after 12 months.
• Steve-001: Skyfall AI launched a system of cooperating AI agents that now operates three small businesses the startup acquired. One business reportedly increased monthly revenue by 127.7%, while another remained 3% below its purchase level. Human owners still approve major decisions.
• ChatGPT Enterprise on AWS Marketplace: OpenAI made its enterprise ChatGPT subscription available through AWS Marketplace, so eligible organizations can purchase the service through existing AWS billing arrangements and apply spending toward AWS commitments.
• Dageno AI 2.0: Dageno helps brands see whether AI assistants recommend their products or competitors and identifies sources influencing those answers. Its free market lookups cover more than 12,000 niche markets, with paid options for monitoring custom prompts. The updated product launched on Product Hunt, the software and product discovery platform.

🏢 Big Tech & Major Companies

• Paramount Skydance CEO David Ellison predicted that AI could transform entertainment, comparing its potential impact on filmmaking to computer-generated imagery. He said the technology would substantially change production while maintaining that artists would remain essential to creative work.

🤖 AI Agents & Infrastructure

• OpenAI introduced Control Room, an internal system that monitors AI research workloads and can halt model output within 35 seconds after a decision to intervene. It allows researchers to stop experiments when safety or security concerns arise and is designed to stop output if the monitoring system itself fails.

🏛️ AI Policy, Governance & Safety

• AI safety organizations outlined their post-election plans for Washington. Anthropic-linked Public First Action and affiliated political groups want to support candidates who favor AI safeguards and oppose federal legislation that overrides state protections. The groups reportedly have $30 million available for their efforts.

💡 Industry Commentary & Analysis

• An anti-Waymo campaign used newspaper opinion pieces flagged for substantial AI involvement, according to Semafor. AI detection company Pangram identified significant AI-generated content in several pieces criticizing autonomous taxis, while the newspapers did not disclose AI assistance. Detector scores alone cannot establish the exact amount of AI-written text.

Previous Around the Horn Digests

Catch up on the stories you missed:

• Monday, November 9, 2026: Anthropic's $3.4 billion power deal, Trump's openness to AI safeguards, and coding agents struggling with real-world tasks.
• Friday, November 6, 2026: Anthropic's accelerating AI research, OpenAI's cybersecurity model, and Claude's access to financial accounts.
• Thursday, November 5, 2026: AI agents discovered Linux vulnerabilities, researchers investigated mysterious cell particles, and election-related AI lobbying faced voters.
• Wednesday, November 4, 2026: OpenAI's security concerns, Google's school AI policies, and Apple's warnings about autonomous agents.
• Tuesday, November 3, 2026: Anthropic's AI acceleration warnings, deceptive deepfake apps, and election misinformation preparations.
• Monday, November 2, 2026: OpenAI's ChatGPT and Codex changes, Meta's advertising growth, and Apple's China rollout plans.
• Friday, October 30, 2026: AI monitoring tests, Reddit's search traffic concerns, and Mistral's massive open model.

That's a Wrap

That's 22 AI news and product links to get through today. If you made it all the way here, congratulations. You've now survived an IPO email correction, an AI security incident, and a company that put agents in charge of three businesses.

At least nobody asked you to approve the agents' expense reports.

For the shorter daily version, subscribe to The Neuron. We sort through the news so you can catch up in a few minutes.

See you tomorrow.

P.S. Know someone who needs an AI news catch-up? Send them this roundup and tell them to subscribe here.