Andrej Karpathy Says AGI Decades Away, Calls AI Agents 'Slops'
Karpathy’s comments come amid a frenzy of AI agent launches, with companies touting autonomous digital assistants that can browse, code, and execute complex workflows.

Andrej Karpathy — one of the most respected figures in modern artificial intelligence and a key architect behind OpenAI’s GPT models and Tesla’s Autopilot — is not buying the current AI hype. In a new episode of The Dwarkesh Podcast, Karpathy said that Artificial General Intelligence (AGI) is still “decades away” and dismissed the current wave of AI agent startups as overblown.
“Overall, the models, they are not there,” Karpathy said. “And I feel like the industry [...] it's making too big of a jump and it's trying to pretend that this is amazing. And it's not — it's slop! And I think they are not coming to terms with it. Maybe they are trying to fundraise or something like that, I’m not sure what’s going on.”
Karpathy’s comments come amid a frenzy of AI agent launches, with companies touting autonomous digital assistants that can browse, code, and execute complex workflows. But for the former OpenAI founding member and ex-Senior Director of AI at Tesla, the reality is far from science fiction.
The @karpathy interview
— Dwarkesh Patel (@dwarkesh_sp) October 17, 2025
0:00:00 – AGI is still a decade away
0:30:33 – LLM cognitive deficits
0:40:53 – RL is terrible
0:50:26 – How do humans learn?
1:07:13 – AGI will blend into 2% GDP growth
1:18:24 – ASI
1:33:38 – Evolution of intelligence & culture
1:43:43 - Why self… pic.twitter.com/rpVYjKNTNZ
He was also the inventor of the term “vibe coding,” which means guiding AI models to write software through natural language prompts — focusing more on describing what you want built rather than manually coding how to build it.
Recently, he launched a lightweight $100 ChatGPT alternative. When he was asked if he leveraged any AI coding models like Claude or GitHub Copilot, Karpathy said the entire codebase is handwritten.
He mentions he did try to use Claude/ Codes agents, but they were 'unhelpful'. His candid remarks reflect a growing unease among AI veterans who believe the field may be running ahead of its technical limits — and that the road to true AGI might be far longer, and messier, than Silicon Valley’s pitch decks suggest.
Interestingly, AI scientist and long-time critic of generative AI, Gary Marcus, has been voicing similar concerns for years — warning that today’s AI models are overhyped, unreliable, and far from truly intelligent. Reacting to Karpathy’s remarks, Marcus quipped on X, “When Karpathy sounds like Marcus, the gig is up.”
Comments ()