Anthropic hires OpenAI co-founder Andrej Karpathy to lead Claude pre-training research
Anthropic scored a major hire today. Former Tesla senior director and OpenAI founding member Andrej Karpathy is joining the organization as a member of the company’s pre-training team.
Karpathy posted on X (formerly Twitter) this Tuesday and said, “Personal update: I’ve joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.”
Where has Karpathy worked?
After completing his Stanford University Ph.D. in 2016 (studies which he detailed as focused on novel convolutional/recurrent neural networks and their applications in computer vision, natural language processing, and their intersection) and having previously worked as a research intern on the Google DeepMind team, Karpathy was an initial founding member of OpenAI, holding the role of research scientist from Jan 2016 to Jun 2017.
After leaving OpenAI, Karpathy worked at Tesla as senior director for AI for a period of just over five years. His work at Musk’s automotive manufacturing and development company saw him lead the computer vision team of Tesla Autopilot.
The feeling within Anthropic, it appears from these hires — given Karpathy’s large-scale training prowess, is that research and development driven by AI-accelerated functions is more of a competitive differentiator than core compute muscle or even capacity alone.
Anthropic collects OpenAI employees
Closing off his previous employment tenure date immediately, Karpathy will work for Nicholas Joseph, another ex-OpenAI employee who left for Anthropic after just nine months with the rival AI model pioneer.
“Excited to welcome Andrej to the Pretraining team! He’ll be building a team focused on using Claude to accelerate pretraining research itself. I can’t think of anyone better suited to do it – looking forward to what we build together!” wrote Joseph, on X.
Alongside Joseph, Karpathy will share water cooler chats with AI luminary John Schulman, also previously part of the OpenAI co-founder group and a man whose own blog lists his core interests in robotics and reinforcement learning.
OpenAI has also hemorrhaged its chief scientist, Ilya Sutskever (now chief scientist at Safe Superintelligence Inc) and its former CTO Mira Murati (now co-founder and CEO at Thinking Machines Lab) in the last couple of years.
Why Claude pre-training matters
A key focus for Anthropic, Claude pre-training work sees the company “feed” its foundation model with a variety of diverse datasets that span text, audio and visual media and software code in order to build model pattern knowledge.
“This is like a leading franchise recruiting someone who’s simultaneously the best player & the league’s best broadcaster & its most watched developmental coach all in one.” – Tech commentator @signulll.
Claude’s pre-training starts from a foundational cornerstone known as the Claude Constitution. Anthropic has described this mandate and said that, “Claude’s constitution is a detailed description of Anthropic’s intentions for Claude’s values and behavior.”
Pseudonymous tech commentator @signulll posted to their 198.2K X followers this Tuesday and said of Karpathy’s appointment, “This is like a leading franchise recruiting someone who’s simultaneously the best player & the league’s best broadcaster & its most watched developmental coach all in one.”
The man who coined vibe coding
A prominent player in the AI space for the last decade, Karpathy is the person who coined the term vibe coding back in February last year.
He said at the time, “There’s a new kind of coding I call ‘vibe coding’, where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It’s possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good.”
Anthropic boosts security leadership
Anthropic’s brain power dominance is further bolstered this month with the appointment os security software engineer Chris Rohlf, who joins the organization’s frontier red team.
“The speed of AI progress is astounding. We have a real opportunity in front of us to dramatically improve cybersecurity with AI. I can’t think of a better company or team to join at this critical moment in time,” posted Rohlf.
Dedicated to proactively working on AI model cybersecurity vulnerability protection through stress testing, Anthropic has said its Frontier Red Team enhances its ability to advance the frontier of AI rapidly and with the confidence that it are doing so responsibly.
The foundation model race continues
With Anthropic, OpenAI and Google constantly vying for a position of leadership in the hearts and minds of developers and end users alike, the companies will naturally take major staff movements as a sign of strength or weakness in corresponding measure.
The feeling within Anthropic, it appears from these hires — given Karpathy’s large-scale training prowess, is that research and development driven by AI-accelerated functions is more of a competitive differentiator than core compute muscle or even capacity alone.
Karpathy writes his own Neural Networks: Zero to Hero blog and hosts his own YouTube channel.
