Karpathy Joins Anthropic Pretraining Team to Use Claude for Research Acceleration
Andrej Karpathy — co-founder of OpenAI and former Tesla AI Director — has joined Anthropic specifically within the Pretraining team. His mandate, as announced by Pretraining lead Nicholas Joseph, is to build a dedicated team focused on using Claude itself to accelerate Anthropic's pretraining research — a model-accelerates-its-own-training feedback loop. Karpathy describes the next few years at the frontier of LLMs as "especially formative." His public work on context engineering, LLM wikis, and autonomous research loops maps directly onto Anthropic's Claude Code and managed-agents infrastructure.
Why It Matters
Placing Karpathy specifically in Pretraining — to use the model being built for accelerating its own training — is a structural closing of the AI self-improvement loop within a safety-focused lab. This is qualitatively different from a general researcher role: it signals Anthropic's intent to systematically compress the pretraining research cycle using frontier AI capability.