OpenAI Co-Founder Breaks Free from AI Tools, Crafts Hand-Coded Chatbot Model

OpenAI Co-Founder Ditches AI Tools for Hand-Coded Chatbot Model Andrej Karpathy, co-founder of OpenAI, has released an open-source large language model (LLM) called nanochat, marking a departure from his previous stance on "vibe coding." In a surprise move, Karpathy revealed that he wrote the entire 8,000-line codebase by hand, citing difficulties with AI tools. According to Karpathy's blog post, nanochat is designed to be a minimal, full-stack training-inference pipeline for building ChatGPT-style chatbots. The project allows users to create their own LLMs in hours and at a cost of as little as $100. "It's basically entirely hand-written (with tab autocomplete)," Karpathy wrote. Karpathy's decision to abandon AI tools, which he had previously championed, raises questions about the limitations of current AI technology. In an interview with Gizmodo, Karpathy explained that he tried using Claude Codex agents but found them unhelpful. "I tried to use Claude Codex agents a few times but they just didn't work well enough at all and net unhelpful," he said. The release of nanochat comes over a year after Karpathy's departure from OpenAI, where he was instrumental in developing the company's LLMs. His decision to write the code by hand has sparked debate among AI researchers and developers about the role of human creativity in AI development. Background Karpathy popularized the term "vibe coding" last year, describing it as a practice where developers rely on AI tools to generate code. However, his own experience with nanochat suggests that there may be limitations to this approach. As Karpathy noted, writing code by hand can be a more efficient and effective way to achieve specific goals. Implications The release of nanochat has significant implications for the development of AI technology. It highlights the need for more nuanced understanding of AI capabilities and limitations. "This is a wake-up call for the AI community," said Dr. Emily M. Bender, a leading researcher in natural language processing. "We need to be more aware of what our tools can and cannot do." Next Developments The release of nanochat marks the beginning of a new era in LLM development. As researchers and developers explore the possibilities of hand-coded AI models, we may see a shift away from reliance on AI tools. Karpathy's decision to write code by hand has sparked a renewed interest in human creativity and its role in AI development. In conclusion, the release of nanochat by OpenAI co-founder Andrej Karpathy marks a significant departure from his previous stance on "vibe coding." As researchers and developers explore the possibilities of hand-coded AI models, we may see a shift towards more nuanced understanding of AI capabilities and limitations. *Reporting by Developers.*