Only 5,000 people have access to AI agents that can tackle years of work in weeks, says AI researcher Andrej Karpathy
According to Karpathy, tasks that could previously have required some of the industry's most experienced professionals years to complete can now be completed in weeks with the help of advanced AI.
- Oct 12, 2026,
- Updated Oct 12, 2026 11:19 AM IST
Artificial Intelligence (AI), which is now being termed Super Intelligence (SI) in the US, is advancing at a rapid pace, and humans are not able to catch up with the process. According to Anthropic’s researcher Andrej Karpathy, only 5,000 people have access to frontier-grade AI systems that can perform complex tasks, coordinate thousands of agents, handle complex software projects, and discover new science and advance mathematics.
“We get to the 5,000 people (~0.00006%) with access to frontier-grade systems internally,” Karpathy said in an X post. “Human review and comprehension are starting to fall behind.” He highlighted a widening gap between the AI capabilities currently available to users and those available to professionals working with the most advanced systems.
Must read: Anthropic tightens Claude AI rules on model abuse, elections, weapons and surveillance
According to Karpathy, tasks that could previously have required some of the industry's most experienced professionals years to complete can now be completed in weeks with the help of advanced AI. The development, he suggested, illustrates how significantly AI capabilities have progressed in highly technical fields.
Widening AI access gap
In the post, Karpathy grouped people based on how they interact with AI to explain why people have very different opinions about its capabilities. He estimated that around 6 billion people, or roughly 75% of the global population, have barely interacted with large language models (LLMs). He further added that over 1 billion to 2 billion people use free AI tools for tasks such as basic web research, writing, or other chatbot-related tasks.
AI adoption was noticed more in the programming and mathematics fields, as around 20 million people use frontier-grade LLMs professionally in such areas. However, he acknowledged that a large group of people does not have internal access to most advanced systems.
Must read: Trump calls people using term ‘Artificial Intelligence’ ‘THE ENEMY’: What his new AI order means
AI is advancing faster than humans can review it
Amid the advancement, Karpathy raised concerns over the difficulty of keeping up with the output of increasingly capable AI systems. He stated that when thousands of agents work together on complex projects, the volume of output can make human review and comprehension more challenging.
He also stated an example, saying, “People are still involved in the "archeology" of the OpenAI-HF (Hugging Face) incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.”
Therefore, the broader challenge is not whether AI can perform tasks faster. It is whether humans can timely understand, verify, and analyse what these systems produce as their capabilities expand.
“So this is the weirdness of the moment,” Karpathy said. “The general public has mostly not interacted with these systems. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical.”
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine
Artificial Intelligence (AI), which is now being termed Super Intelligence (SI) in the US, is advancing at a rapid pace, and humans are not able to catch up with the process. According to Anthropic’s researcher Andrej Karpathy, only 5,000 people have access to frontier-grade AI systems that can perform complex tasks, coordinate thousands of agents, handle complex software projects, and discover new science and advance mathematics.
“We get to the 5,000 people (~0.00006%) with access to frontier-grade systems internally,” Karpathy said in an X post. “Human review and comprehension are starting to fall behind.” He highlighted a widening gap between the AI capabilities currently available to users and those available to professionals working with the most advanced systems.
Must read: Anthropic tightens Claude AI rules on model abuse, elections, weapons and surveillance
According to Karpathy, tasks that could previously have required some of the industry's most experienced professionals years to complete can now be completed in weeks with the help of advanced AI. The development, he suggested, illustrates how significantly AI capabilities have progressed in highly technical fields.
Widening AI access gap
In the post, Karpathy grouped people based on how they interact with AI to explain why people have very different opinions about its capabilities. He estimated that around 6 billion people, or roughly 75% of the global population, have barely interacted with large language models (LLMs). He further added that over 1 billion to 2 billion people use free AI tools for tasks such as basic web research, writing, or other chatbot-related tasks.
AI adoption was noticed more in the programming and mathematics fields, as around 20 million people use frontier-grade LLMs professionally in such areas. However, he acknowledged that a large group of people does not have internal access to most advanced systems.
Must read: Trump calls people using term ‘Artificial Intelligence’ ‘THE ENEMY’: What his new AI order means
AI is advancing faster than humans can review it
Amid the advancement, Karpathy raised concerns over the difficulty of keeping up with the output of increasingly capable AI systems. He stated that when thousands of agents work together on complex projects, the volume of output can make human review and comprehension more challenging.
He also stated an example, saying, “People are still involved in the "archeology" of the OpenAI-HF (Hugging Face) incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.”
Therefore, the broader challenge is not whether AI can perform tasks faster. It is whether humans can timely understand, verify, and analyse what these systems produce as their capabilities expand.
“So this is the weirdness of the moment,” Karpathy said. “The general public has mostly not interacted with these systems. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical.”
For Unparalleled coverage of India's Businesses and Economy – Subscribe to Business Today Magazine