What is the true meaning of OpenAI's co-founder's view that 'it will take another 10 years to realize AGI'?

Andrej Karpathy , a Slovak-Canadian computer scientist who served as director of AI and Autopilot Vision at Tesla and co-founder of OpenAI, the developer of ChatGPT, appeared on the Dwarkesh Podcast, a podcast where researchers ask questions, and said that it will take 10 years for AI agents to become fully functional, and that it will take at least that long to achieve AGI (artificial general intelligence).

Andrej Karpathy — AGI is still a decade away
https://www.dwarkesh.com/p/andrej-karpathy

AI researcher Andrej Karpathy says AGI is still a decade away, here's why | Technology News - The Indian Express
https://indianexpress.com/article/technology/artificial-intelligence/andrej-karpathy-agi-is-still-a-decade-away-heres-why-10316244/

Karpathy has long argued that while people in the AI industry say 2025 will be the year of AI agents, he believes 2025 to 2035 will be the decade of AI agents. He argues that current AI agents often don't function well, and that a huge amount of work across a wide range of fields will be required before they can function properly as AI agents.

Projects like OpenAI's Operator are to the digital world as Humanoid robots are to the physical world. One general setting (monitor keyboard and mouse, or human body) that can in principle gradually perform arbitrarily general tasks, via an I/O interface originally designed for… https://t.co/pZctgBa6PV

— Andrej Karpathy (@karpathy) January 23, 2025

During a podcast appearance, Karpathy was asked about the claim that '2025 is not the year of AI agents, but 2025-2035 will be the decade of AI agents.' Karpathy replied, 'I don't remember exactly who said it, but they were implying that this year will be the year of AI agents in terms of large-scale language models and their future evolution. I was inspired by this statement because there is a lot of over-prediction going on in the AI industry.'

Karpathy himself has used AI agents such as Anthropic's Claude and OpenAI's Codex, and acknowledges that they have excellent performance, but he feels there is still much work to be done. He also predicts that we will continue to use AI agents for the next 10 years, and that it will be some time before we reach AGI.

Karpathy pointed out that existing AI intelligence is insufficient and cannot handle multimodal behavior, including computer operations. 'They (AI agents) do not have the ability to continuously learn. They cannot remember something just by being taught it. They lack cognitive abilities and do not function well. It will take 10 years to solve all of these problems,' he said.

After the podcast was released, Karpathy provided additional explanations about the content in a post on X.

My pleasure to come on Dwarkesh last week, I thought the questions and conversation were really good.

I re-watched the pod just now too. First of all, yes I know, and I'm sorry that I speak so fast :). It's to my detriment because sometimes my speaking thread out-executes my… https://t.co/bnPSrY74px

— Andrej Karpathy (@karpathy) October 18, 2025

Karpathy argues that his AI timeline is 5-10 times more pessimistic than what you'd see in AI industry house parties or X timelines, but is still significantly more optimistic than AI deniers and skeptics. 'Overall, I think the 10-year timeline is a very bullish outlook for AGI. It just doesn't feel that way in comparison to the current hype,' he said.

'We need AI agents that can collaborate with humans, not ones that can churn out 1,000 lines of code in 20 minutes, leaving humans no time to evaluate it,' Karpathy said. 'We need to be more realistic about the capabilities of AI tools and their place in the modern industry. If we don't do this properly, I'm concerned that we'll see a rise in poorly written code across the software landscape, leading to increased vulnerabilities and security breaches.'