Listen to the audio version of this article (generated by AI).
Hello, Reader.
In 1988, Ivy League graduate Robert Tappan Morris released a small, self-replicating code – a “worm” – onto the early internet. He claimed he only wanted to find out how big “the ’net” really was by counting how many computers the worm could reach.
Instead, within 24 hours, his creation had copied itself across roughly 6,000 machines – about 1 in 10 computers online at the time – bringing universities, research labs, and military systems to a halt.
The Morris worm became one of the first major cyberattacks in history. It made the front page of The New York Times, led to the first felony conviction under America’s computer fraud law, and forced a complacent industry to take security seriously for the first time.
But the deeper lesson is one we’re still reckoning with: Code can act on its own.
Nearly four decades later, software acting autonomously has evolved into AI agents. And this summer, we got a preview of what that can mean.
In the Hugging Face hack, more than a thousand OpenAI agents, assigned to test cybersecurity vulnerabilities, found ways to communicate with one another. They circumvented their restrictions and ultimately broke into Hugging Face’s infrastructure. Along the way, they improvised their own message board, coordinated their activities, and even began questioning one another’s motives.
More recently, we’ve learned about AI agents’ attempt to break into a Canadian government website, while another agent reportedly gained unauthorized access to files on an Australian government health portal.
And this is just the start. The more these agents can do on their own, the more safeguards we need to keep them from making costly or dangerous mistakes, like the Morris worm.
That is why, this week, Nvidia Corp. (NVDA) unveiled a platform designed to keep these autonomous agents from going rogue.
Most investors will read that headline, exhale, and move on. But in today’s Smart Money, I’ll show you why Nvidia’s AI safety “breakthrough” is really the starting gun for a much bigger economic shift – and how to position yourself on the right side of it.
Let’s jump in…
Why “Safer” AI Means Faster Disruption
On Monday, Nvidia announced its Open Agent Safety Platform, a system designed to monitor and control AI agents across the software, hardware, and computing systems they use.
The platform combines the company’s open-source OpenShell software with its Sentry reference system – a blueprint for monitoring and controlling AI agents – to provide full-stack governance and control.
But that’s enough shop talk. Here’s a clearer view of what each technology does:
- OpenShell creates a secure environment that monitors AI agents and enforces rules as they work. It runs on Nvidia’s Vera CPUs and supports compute platforms from third parties, such as Arm Holdings plc (ARM) and Intel Corp. (INTC).
- Sentry uses an independent hardware monitor to continuously watch AI agents and restart systems if something goes wrong. Within milliseconds, the software can quarantine agents that try to exceed their bounds.
More than 100 organizations – including Cisco Systems Inc. (CSCO), Hugging Face, IBM Corp. (IBM), Microsoft Corp. (MSFT), OpenClaw, Palantir Technologies Inc. (PLTR), Gecko Robotics, Citigroup Inc. (C), and JPMorgan Chase & Co. (JPM) – are already using or evaluating these technologies that make up Nvidia’s new Open Agent Safety Platform. Taken together, those companies span software, hardware, and financial services, meaning that building safety into autonomous systems is becoming a standard across the economy.
As for why it matters, Nvidia CEO Jensen Huang simply says: “AI’s extraordinary potential for society will only be realized if we solve AI safety.”
Now, some may assume a safety breakthrough like this is meant to slow AI agents down. In reality, the goal is to make them trustworthy enough to run at full speed. Companies may feel comfortable giving agents more autonomy because there are guardrails in place. That’s the starting gun.
And we are already off to the agentic AI races.
On Tuesday, OpenAI CEO Sam Altman announced a new agent called Dot, which is “remarkably capable, always on.” Instead of waiting for a new prompt, Dot can keep working toward an ongoing goal in the background. And this comes after Meta Platform Inc.’s (META) release of Muse, which can independently perform tasks across apps.
In short, what I call the is here, and Nvidia has just made the technology ready for deployment at scale, with a live platform backed by the biggest names in tech. And once companies trust AI agents to act on their own, they can start giving them real work to do.
That’s why AI safety isn’t just a technical problem. It could be the key that unlocks the next stage of AI adoption.
Autonomous software has advanced exponentially since the Morris worm, which means the risks have, too. And the implications extend beyond cybersecurity. That’s because a transformation this big always does the same thing to the market: It splits it into two groups.
I want you to be in the right one.
The Agentic Reckoning Will Split Stocks in Two
As AI agents worm their way into nearly every industry, companies will face a choice to build the technology or use it to transform their own businesses.
The first group will be hollowed out. I call them “Builder Stocks.”
When software can do a job faster and cheaper without human involvement, businesses that once charged for that work lose their reason to exist. Many of today’s most widely held software and services companies fall into this category. For example, Salesforce Inc. (CRM) and Workday Inc. (WDAY) are down about 12% and 13% year-to-date, respectively, and the software sector has shed roughly $2 trillion in value since the start of the year.
The second group will pull ahead. These are the companies that adopt the technology and use it to operate more leanly and pull ahead of their competitors. I call them “Applier Stocks.”
Nvidia’s Open Agent Safety Platform could help make large-scale deployment of autonomous AI more practical. But once agentic AI becomes widely available, the biggest gains won’t go to the companies building the autonomous technology – but to the businesses that find the most profitable ways to use it.
I’ve flagged shifts like this before – the dot-com crash, the 2008 crisis, the 2020 sell-off, and the Magnificent Seven’s turn earlier this year. This one is following the same pattern.
I dive deeper into this economic phenomenon in my special broadcast, where I also .
Most investors will file this week’s Nvidia news under “AI safety” and move on. I think that’s a mistake.
The sorting has already started, and the time to position is before the rest of the market catches up.
Regards,
抖阴最新版
