OpenAI invested roughly $1.2 million in API costs to produce functional exploits targeting 948 Linux kernel vulnerabilities. During remarks at The New York Times' DealBook Summit on Wednesday, CEO Sam Altman cited this research as proof that developers can tackle challenging engineering problems by allocating significantly greater computational resources.

Altman, joined by OpenAI President Greg Brockman, suggested the approach extends beyond security research. "Apply that to optimizing your code, apply that to solving other hard problems. All that's coming," Altman stated.

The Linux work sits within OpenAI's wider push into compute-intensive AI development, encompassing proprietary silicon, infrastructure buildout, and an agent platform Altman wants to position as the standard for developers. Altman views available compute capacity as the fundamental constraint on what these systems can ultimately achieve.

Hunting kernel bugs at scale

OpenAI's November 4 technical report, titled "A thousand fewer paths to root," describes an automated pipeline that reproduced 1,724 vulnerabilities and generated working exploits for more than half. The agents not only identified flaws but also reproduced them, tested exploitability, and authored patches. Human security researchers validated the work before sharing findings with Linux kernel maintainers, though OpenAI's figures do not confirm all 948 vulnerabilities have undergone independent verification or received fixes.

The $1.2 million figure represents a ceiling on costs, according to the company, since the team had not yet refined the harness for cost optimization. The outcome illustrates the scale of work agents can complete when furnished with substantial inference budgets. OpenAI is also constructing long-running agents capable of consuming thousands of dollars in compute daily.

OpenAI wants to own the AI stack

Altman articulated a vision of vertical integration across the AI supply chain. "I think we are set up to be like vertically integrated, to deliver the best thing across all the ways people need to use it, and we're thrilled for people to build very successful businesses on top of us," he said.

Over recent months, OpenAI has expanded investments spanning custom silicon, computing infrastructure, frontier models, and an agent platform targeting both consumer and enterprise segments. The company has also moved deeper into the stack by developing its own silicon.

We are the only company that has the entire intelligence stack.

Sam Altman

A custom chip for inference

OpenAI's Jalapeño inference chip is currently in development, with preliminary benchmarks indicating improvements in throughput per watt and token latency. At DealBook, Altman made an ambitious assertion about its capabilities.

We have our own chip coming, which is I think going to be the best inference chip in the industry.

Sam Altman

The company is also working to consolidate its software offerings. "My focus and our focus on product's been to have one agentic platform that you're getting everyone to use," Altman explained. He did not unveil modifications to OpenAI's developer APIs or announce restrictions affecting third-party developers, and provided minimal technical specifics regarding the proposed platform's mechanics.

Compute is the real limiter

Altman contends that AI models will become indispensable for business operations, yet the extent of what organizations accomplish will hinge on their access to compute resources. Recent comparisons of coding-agent frameworks have demonstrated how execution expenses fluctuate based on the software layer surrounding a model.

I think that the economy again is going to be massively accelerated through this technology. I think that the biggest limiter really just will be compute.

Sam Altman