Version 2

Andrej Karpathy, co-founder of OpenAI and former Tesla AI director, announced on May 19, 2026 that he has joined Anthropic [1]. Coverage from TechCrunch, CNBC, Axios, the WSJ, and Forbes confirmed within hours that his role is specifically on Anthropic's pre-training team [2][4][5][6]. The hire is arriving alongside a broader Anthropic momentum story: the company recently closed a $30B fundraise at a reported $900B valuation [18][19], and at least one observer noted it coincided with Anthropic's first profitable quarter [20].

Karpathy is among the most credible and publicly influential figures in AI research, and his choice to focus specifically on pre-training at Anthropic is a signal about where frontier capability work is headed. The competitive framing has been immediate: commentators are treating the hire as a direct symbolic loss for OpenAI, the organization Karpathy co-founded [15][14]. It also raises questions about the future of his AI education work, which had been his primary public identity since leaving OpenAI.

On May 19, 2026, Andrej Karpathy posted to Twitter that he had joined Anthropic [1], setting off a wave of coverage from nearly every major technology outlet. TechCrunch, CNBC, Axios, the Wall Street Journal, and Forbes all confirmed the hire within hours [2][3][4][5][6]. Crucially, the reporting added a specific detail absent from Karpathy's own announcement: he has joined Anthropic's pre-training team, the group responsible for the foundational training of Claude models [2][7][8][9].

Karpathy's career arc makes the hire notable. He was a co-founder of OpenAI, later left to lead AI at Tesla, and then returned to OpenAI before departing again in early 2024. After that exit, he announced Eureka Labs, an AI-native education company [10][11], and spent roughly two years producing widely watched AI education content. Approximately one month before the Anthropic announcement, he had publicly mused about the appeal of joining a major frontier lab — noting that such organizations "are kind of like at the edge of capability of what's possible" [12]. His Anthropic announcement echoed that framing, describing the current LLM moment as "especially formative" [13], and acknowledged his ongoing passion for education while flagging that it would be deprioritized for the near term.

The community reaction framed the move primarily as a competitive signal. Commentators positioned it as a meaningful win for Anthropic and a symbolic loss for OpenAI, with one widely retweeted post describing Karpathy as "OpenAI's 11th co-founder" now working for a rival [14][15]. Some ironic meta-commentary circulated noting the whipsaw news cycle: Karpathy's move to Anthropic was immediately followed by OpenAI announcements that shifted attention back, prompting jokes about the volatility of AI-competitive narratives [16][17]. The hire also landed in the context of Anthropic's recently reported $30B fundraise at a $900B valuation [18][19], with at least one observer noting it coincided with the company's first profitable quarter [20].

One fringe claim circulating on social media asserted that Karpathy joined specifically to lead "recursive self-improvement" research [21]. This has not been corroborated by any outlet and appears to be speculation beyond what the confirmed pre-training team assignment implies. The more grounded open question is what Karpathy's specific mandate within pre-training will be — whether that centers on scaling methodology, data strategy, architectural research, or direct model improvement work — a detail that neither Karpathy nor Anthropic has made public.