Andrej Karpathy Joins Anthropic to Build Claude-Led Pre-Training Team After 2 OpenAI Stints
Updated
Updated · TechCrunch · May 19
Andrej Karpathy Joins Anthropic to Build Claude-Led Pre-Training Team After 2 OpenAI Stints
6 articles · Updated · TechCrunch · May 19
Karpathy said Tuesday he started at Anthropic this week and is building a team that uses Claude to speed pre-training research under team lead Nick Joseph.
Pre-training drives the large-scale runs that give Claude its core capabilities and is among the most compute-intensive, expensive stages of developing frontier AI models.
Anthropic’s hire signals a bet that AI-assisted research can sharpen its edge against OpenAI and Google, not just bigger compute budgets.
Karpathy previously co-founded OpenAI, led Tesla’s Autopilot and Full Self-Driving efforts until 2022, then returned to OpenAI for a year before leaving in 2024 to start Eureka Labs.
Anthropic also added cybersecurity veteran Chris Rohlf to its frontier red team, expanding work on stress-testing advanced models against severe threats.
Karpathy's team will accelerate AI research, but can Anthropic stop its own AI from creating new, unforeseen security risks?
As AI automates its own R&D, can safety research ever win the race against the quest for market dominance?
The 2026 AI Talent War: Andrej Karpathy’s Anthropic Move and the Future of Self-Improving LLMs
Overview
Andrej Karpathy, a leading AI expert with experience from OpenAI and Tesla, officially joined Anthropic on May 19, 2026. His move is a major win for Anthropic, highlighting the fierce competition for top AI talent and the growing importance of expertise in generative AI. At Anthropic, Karpathy will lead a team focused on pre-training research, using the Claude model to push the boundaries of large language models. He believes the next few years will be especially formative for LLMs and is excited to return to hands-on research and development.