Anthropic has hired Andrej Karpathy, the researcher who co-founded OpenAI and led Tesla’s autonomous driving programme, to head a new team using Claude to accelerate pre-training research. Karpathy, who joined this week, said he regards the coming years at the frontier of large language models as especially formative and expressed enthusiasm about returning to core research and development.
Working under pre-training lead Nick Joseph, Karpathy will focus on one of the most compute-intensive phases of building frontier models — the large-scale training runs that underpin Claude’s core capabilities. Despite his previous education startup Eureka Labs, Karpathy said he intended to return to that work in time.
Separately, cybersecurity veteran Chris Rohlf, formerly of Meta and Yahoo’s elite security team, has joined Anthropic’s frontier red team, which stress-tests advanced AI models against severe threats. Rohlf said he saw a significant opportunity to improve cybersecurity through AI.
