Andrej Karpathy, a key figure behind OpenAI’s rise, hasn’t written a line of code in months. Instead, he’s been relying heavily on AI agents — even calling his experience a “state of psychosis.”
From Coder to Overseer: A Radical Shift in Workflow
Karpathy’s confession marks a striking change in how software development works today. Once hands-on with 80% of his code, he now delegates most of it to AI agents. Since December, he estimates he’s done just 20% of the coding himself.
"I don’t think a normal person actually realizes that this happened or how dramatic it was," he said during a recent podcast interview. This shift isn’t just personal — it’s becoming the norm across the tech world as agentic AI tools like OpenClaw reshape how work gets done.
OpenClaw, which Karpathy uses extensively, connects to multiple apps and services. It handles everything from managing calendars to sending messages, making it a powerful assistant for both professional and home tasks. Karpathy’s own home now runs on an AI agent he nicknamed “Dobby the House Elf claw,” controlling lights, security, HVAC, and even alerting him about deliveries.
AI Agents Changing the Game—and Mindsets
Karpathy’s experience highlights the rapid pace at which AI is transforming software development.
But it also reveals a deeper psychological impact on creators who once wrote every line themselves.
He described his current mental state as “psychosis,” trying to grasp what AI can do and pushing those boundaries. "I want to be at the forefront of it, and I’m very antsy that I’m not," he admitted, pointing to the flood of new ideas emerging from others in the AI community.
The tension between human creativity and machine assistance is playing out in real time. Karpathy’s reliance on AI agents means he’s more a conductor than a player now, orchestrating complex systems without manual coding. This makes people wonder about the future role of AI researchers and developers.
OpenAI’s Big Bet on Growth Amid Rising Competition
While Karpathy steps back, OpenAI is gearing up for expansion. The company plans to nearly double its workforce this year, aiming to grow from about 4,500 employees to nearly 8,000, according to the Financial Times. New hires will focus on product development, engineering, research, and sales.
San Francisco’s office space for OpenAI will more than double, reflecting the company’s ambition to keep pace with rivals like Anthropic and Google. The AI race is heating up, with firms vying to attract corporate clients eager to deploy AI tools that automate coding and analyze complex data.
Recent acquisitions by OpenAI, including startups specializing in developer tools and AI security, show a clear strategy to broaden its AI capabilities and ensure safer deployment of AI agents. It’s a high-stakes game where innovation and trust must go hand in hand.
Industry Voices Warn of AI’s Risks and Short-Term Focus
Not everyone shares Karpathy’s excitement. Geoffrey Hinton, known as the "godfather of AI," has sounded alarms about the dangers of AI advancing without proper safeguards. He pointed out that many tech companies prioritize short-term profits over long-term consequences.
Hinton argues that researchers focus on immediate challenges, like improving image recognition or video generation, rather than considering how AI might reshape humanity’s future. He estimates a 10% to 20% chance that superintelligent AI could threaten human existence.
Meanwhile, the misuse of AI by bad actors is already a reality. Cyberattacks and fake videos are part of the growing toolbox of digital threats powered by AI. Anthropic recently disrupted a large-scale AI cyberattack linked to a Chinese state-sponsored group targeting numerous organizations.
Karpathy’s shift from coder to AI overseer captures the tech world’s uneasy transition into an era dominated by intelligent agents. As OpenAI races to expand amid fierce competition, the debate over AI’s risks and rewards only intensifies.
