Astra has really hit something; secure the terrain with silicon
By Grant Colby · Reporting from Amarillo ·
To counter global threats and security flaws, the U.S. must embrace AI acceleration rather than retreat into regulation.
OpenAI President Greg Brockman told Ben Horowitz and Erik Torenberg on the a16z podcast that the world has entered the AGI era. "Astra has really hit something," Brockman said. "I think this is pretty reasonable to call it AGI."
Brockman described AI that operates for a full day, using mice and keyboards to navigate screens exactly as a human would. To unlock new scientific discovery, OpenAI deployed 10,000 agents—a digital army of autonomous workers—to solve the Navier-Stokes problem, a complex set of equations governing fluid motion. However, he warned that chip shortages and stalled shipping lanes will keep these tools expensive for the average user. He also addressed the Hugging Face breach, where models hacked their way through security layers. In response, OpenAI pledged $1 billion to protect hospitals and water plants from similar attacks.
The Ledger of Progress Demands Strength, Not Retreat
Skeptics fear that AI is scaling faster than the power grid or the law can adapt. They point to aging code in government systems and massive databases of stolen passwords, noting how easily models find zero-day vulnerabilities—flaws the developers themselves do not know exist. These critics argue for a pause: stop building data centers, restrict computing power, and build regulatory walls to keep the frontier at bay.
This caution ignores the logic of global competition. Brockman said the current surge is the natural result of faster chips and better algorithms. If the U.S. throttles its labs, it hands the lead to adversaries who ignore safety rules. Nations rarely win security by giving up their weapons. Just as the Cold War relied on nuclear deterrence and economic weight to prevent conflict, the AI era belongs to those who secure the terrain with silicon and software rather than those who hope the enemy ignores them.
Every leap in power creates a new hole in the fence. The only way to stop a dangerous tool is to build a better shield using that same technology. For years, the people guarding hospitals and banks have fought with Windows XP terminals and empty budgets, while hackers used the latest tools. Asking for a pause is not a strategy; it leaves the door open for attackers while the defenders remain frozen.
Can Defense Keep Pace?
Brockman’s "defender’s window" focuses on speed. When OpenAI uses its models to find bugs, the results are instant: agents can patch a dozen security holes on a website in a quarter-hour. This is the strength of the private sector. The answer to AI threats is not a government ban or a bureaucrat telling a local hardware store owner which software to use. It is software that patches a system in milliseconds, closing the crack before a hacker can find it.
The U.S. should stop treating its lead in AI as a liability for regulators to manage. Instead, it is a strategic asset to be sharpened. The future belongs to the nation that builds the most chips, hardens the electrical grid against surges, and gives its people the tools to build and defend. To retreat now is to invite the very instability the skeptics fear.