AI visionary and OpenAI co-founder Andrej Karpathy has officially joined Anthropic. In this new role, Karpathy will lead the pre-training team while simultaneously spearheading a specialized unit dedicated to using the Claude model to accelerate its own iterative development. This move reunites him with several former colleagues and solidifies Anthropic's position in the high-stakes arms race for frontier model supremacy.
Karpathy brings a formidable track record to the table, having previously served as the director of AI at Tesla, where he led the vision team for Autopilot. Recently, he has been an outspoken advocate for unconventional AI testing strategies, coined as:
- Vibe coding: A focus on the intuitive flow and feel of software creation through high-level AI interaction.
- Tokenmaxxing: The aggressive testing of frontier models to push the limits of their output and reasoning capabilities.
Describing himself as being in a state of AI psychosis, Karpathy's experimental energy is expected to drive significant breakthroughs in how models like Claude are trained and utilized for automated software engineering. His hiring signals Anthropic's intent to not just build better models, but to fundamentally transform the pipeline by which these intelligence systems are scaled.
