Editorial illustration for Andrej Karpathy Warns: Agentic AI Remains Far from Industry Promises
Andrej Karpathy Exposes the Reality Behind AI Agent Hype
Andrej Karpathy says agentic AI years from meeting industry hype
Right now, the industry wants you to believe in AI agents, the kind that think and act for themselves. Andrej Karpathy is calling it a fantasy.
He's a founding member of OpenAI and Tesla's former director of AI. He built these things. His skepticism isn't a hunch. It's a diagnosis from inside the lab.
The narrative has been one of imminent revolution. Karpathy sees a long, hard slog ahead. The hype, he suggests, is outpacing the hardware and the code.
Andrej Karpathy, former AI researcher at OpenAI and Tesla, is skeptical of the current hype surrounding agent-based AI and large language models. AI researcher Andrej Karpathy believes that the current euphoria surrounding so-called agent AI is premature. In an interview with podcaster Dwarkesh Patel, Karpathy emphasizes that we should not be talking about a "year of agents" but, more realistically, about a "decade of agents".
Karpathy sees some practical benefits from systems like Codex and Claude Code, but says these models are nowhere near behaving like human interns or skilled employees—the kind of capability many labs are aiming for. "They just don't work" The issues, he argues, are fundamental: the models lack core cognitive abilities, aren't truly multimodal, have no reliable memory, and can't consistently handle complex computer tasks. "They just don't work," Karpathy says.
He estimates it will take around ten years for the field to solve these deep-seated problems, based on his "extrapolation with respect to my own experience in the field." "I feel like the problems are tractable, they’re surmountable, but they’re still difficult. If I just average it out, it just feels like a decade to me," Karpathy says. For now, he sees a disconnect between industry hype and technical reality: "There’s some over-prediction going on in the industry." Karpathy, who left OpenAI in early 2024, is part of a growing group of critics who view today's large language models as useful but fundamentally limited.
Karpathy prefers autocomplete over agent intelligence Karpathy thinks even advanced models like GPT-5 Pro are only useful in narrow roles, such as serving as an "oracle" for code analysis.
That "decade of agents" line isn't a prediction. It's a correction. It deflates the breathless conference talk and venture capital pitches built on a much shorter timeline.
The problems are not bugs. They are foundational. Memory, reasoning, true multimodality.
Karpathy isn't saying these are impossible. He's saying they are hard and distant, a view formed by actually shipping AI products that work. His current preference for simple autocomplete over complex agents tells you everything.
This matters for anyone betting on this technology, financially or professionally. The gap between marketing slides and a system that can reliably replace a human task is a canyon. It will be filled with years of tedious engineering, not a sudden breakthrough.
Karpathy's warning is a plea for patience. The real work is quiet, incremental, and profoundly unglamorous. The hype wants to sell you tomorrow. The builder is telling you to wait ten years.
Further Reading
- Five Trends in AI and Data Science for 2026 - MIT Sloan Management Review
- We're Literally Just Getting Started in AI + B2B. Here's Why 2026 Is When It Gets Real - SaaStr
Related Reading
LLMs & Generative AI ChatGPT's 'Nerdy' tweak rewards goblin metaphors in answers, study finds
LLMs & Generative AI Google tests visual 'magazine-style' UI for Gemini 3 Pro users
LLMs & Generative AI AI Engineers Face Rising Costs, Need New Strategies for Efficiency
From Archives Pathway's '(Baby) Dragon Hatchling' swaps Transformers for neuron-synapse network
From Archives Google lets developers embed live Google Maps data in Gemini AI app outputs
Common Questions Answered
Why does Andrej Karpathy believe the 'year of agents' is actually a 'decade of agents'?
Karpathy argues that current agent-based AI systems are still in their early stages and not yet capable of the complex, human-like task execution that many are promising. His perspective suggests that significant technological development and refinement will be needed before AI agents can truly deliver on their ambitious potential.
What makes Andrej Karpathy's perspective on AI agents credible?
Karpathy brings significant credibility to his AI assessment through his extensive experience working at top-tier tech companies like OpenAI and Tesla. His deep technical expertise and direct involvement in cutting-edge AI research provide him with unique insights into the current limitations and realistic trajectory of agent-based AI systems.
How does Karpathy's view challenge current industry narratives about AI agents?
Karpathy challenges the industry's enthusiastic claims by suggesting that AI agents are far from achieving the level of sophisticated, autonomous task execution that many are predicting. He provides a critical reality check, emphasizing that the development of truly capable AI agents will likely take a decade of continued research and technological advancement.
