The word "quarantine" comes from the Italian for forty days,
Ships once had to wait 40 days offshore before entering port to make sure they weren't carrying anything nasty.
Now NVDIA doing something similar with AI agents — except the suspicious cargo is code.
This week in tech:
NVIDIA released an open platform to fence in AI agents.
Claude Sonnet 5.5 is 30% faster and up to 30% cheaper per task.
Florida wants a court to pause OpenAI's new model work.
In Partnership with Doing AI
What Could AI Fix in Your Business?
Bring us one part of your business that takes too much time, happens too often, or depends too heavily on manual work. We'll audit it, identify where AI can help, and show you what a practical solution could look like.
No generic AI recommendations. Just a clear starting point for your business.
Frontier Reads
Satellite images show the scale of America's data center build-out: The country has nearly 100 hyperscale sites, each drawing at least 100 megawatts at peak. Read more
The Case Against Pacing the AI Frontier: Cognizant's CEO argues against a pause, citing a Hugging Face incident where about 1,200 agents traded over 70,000 messages. Read more
Deep Current
NVIDIA puts AI agents on a leash
NVIDIA launched its Open Agent Safety Platform, combining OpenShell and Sentry to monitor AI agents and shut them down if they wander beyond their allowed boundaries
Sonnet 5.5 Cuts Task Costs
Anthropic released Claude Sonnet 5.5, which runs over 30% faster than Sonnet 5. It uses fewer tokens and tool calls, so a task can cost up to 30% less at the same list price.
Florida Wants OpenAI Paused
Florida's attorney general, James Uthmeier, asked a court to stop OpenAI from building new models until independent safety guardrails exist. OpenAI says it wants industry-wide rules instead.
Starship Is Orbital
SpaceX flew Starship into orbit for the first time and deployed 26 Starlink satellites. It is the first Starship flight to earn revenue.
In Radar
OpenAI ships an assistant while a bigger model waits
OpenAI unveiled "dots," a new AI assistant, even as safety concerns pushed back the release of its next flagship model. The split says something on its own: shipping the smaller, safer thing while holding back the bigger one until it clears internal review. Read more
Anthropic warns: AI could get seriously dangerous
In its IPO filing, Anthropic warned investors that advanced AI could create “catastrophic or existential risks”—including unpredictable behavior and potential real-world harm. Read more
Social Finds
Vibe Coders Burning Tokens
Same Prompt Was Given
Tool Of The Day
Manus 2.0
