AI agents are all the rage right now, but Andrej Karpathy doesn’t seem to believe that they will be as big as some are predicting — at least in the near future.
Former Tesla Director of AI Andrej Karpathy has appeared to disagree with OpenAI’s claims of 2025 being the year of AI agents. OpenAI COO Brad Lightcap had said earlier this month that 2025 would be the year of AI agents. “We now have the ability for enterprises to build agents in a much more visual, much more intuitive way. You’ve heard us say that 2025 has been the year of agents, we think that’s true. Codex has been a great example of us for that,” OpenAI COO Brad Lightcap had said in an interview with Bloomberg.
Karpahty believes that broad AI agent adoption could take as long as a decade as opposed to giving significant years this year. ” I’m not actually sure who said this but they were alluding to this being the year of agents with respect to LLMs and how they were going to evolve,” Karpathy said on the Dwarkesh podcast. “I was triggered by that because there’s some over-prediction going on in the industry. In my mind, this is more accurately described as the decade of agents,” he added.
“We have some very early agents that are extremely impressive and that I use daily — Claude and Codex and so on — but I still feel there’s so much work to be done. My reaction is we’ll be working with these things for a decade. They’re going to get better, and it’s going to be wonderful. I was just reacting to the timelines of the implication,” he added.
Karpathy said that actually making agents work in the real world will take longer than most people expect. He said that current AI agents can’t quite replace full employees. “(AI agents) don’t have enough intelligence, they’re not multimodal enough, they can’t do computer use and all this stuff…they don’t have continual learning. You can’t just tell them something and they’ll remember it. They’re cognitively lacking and it’s just not working. It will take about a decade to work through all of those issues,” he added.
While AI agents seem to have made rapid progress in coding — both OpenAI and Anthropic have said that a large fraction of their code is now written by AI agents — AI agents haven’t quite come into their own in other fields. This could be because coding is a self-contained environment with verifiable results, and many other fields, like writing or analysis can be more subjective, and require more extensive and broad-based computer use. And Andrej Karpathy seems to be saying that it might be as long as 10 years before AI agents are good enough to be able to replace human employees in a wide variety of fields.