Karpathy Calls It: The Death of Manual Programming

Andrej Karpathy declares the manual programming era over, as AI agents rapidly transform how tasks are tackled. December 2025 marked a turning point shift.
Andrej Karpathy, a former OpenAI researcher, recently declared the era of manual programming over. With AI agents now capable of handling complex tasks in minutes instead of days, the programming landscape has indeed become unrecognizable. This change isn't just a trend, it's a seismic shift in how we approach software development.
December 2025: A Turning Point
Karpathy's perspective shifted dramatically by December 2025. He had previously viewed AI agents with skepticism, but December marked a clear transition. Why the sudden change? The capabilities of AI agents reached a level of effectiveness that disrupted traditional programming methods. What once required meticulous coding is now accomplished through smarter, faster AI-driven processes.
Reinforcement Learning: Not a Silver Bullet
Karpathy did address reinforcement learning from human feedback (RLHF), noting its limitations. While it's a useful tool, relying solely on RLHF isn't enough to achieve the desired results in training AI language models. This is a critical point for developers, who need to blend multiple approaches to optimize AI training. The SDK handles this in three lines now, but understanding its limitations remains key.
Why This Matters to Developers
For developers, this shift isn't just academic. It reshapes how software is built, deployed, and maintained. The knowledge gap between manual programming and AI-driven development is widening. Will developers adapt, or risk becoming obsolete? This is the pressing question as AI continues to evolve at a breakneck pace.
As we ship into 2026, the call is clear: embrace AI's role in development or be left behind. It’s not simply about keeping up with trends, but about rethinking the very fabric of programming. Clone the repo. Run the test. Then form an opinion.
Get AI news in your inbox
Daily digest of what matters in AI.
Key Terms Explained
The AI company behind ChatGPT, GPT-4, DALL-E, and Whisper.
A learning approach where an agent learns by interacting with an environment and receiving rewards or penalties.
Reinforcement Learning from Human Feedback.
The process of teaching an AI model by exposing it to data and adjusting its parameters to minimize errors.