OpenAI co-founder Andrej Karpathy has joined Anthropic's pre-training team, where he will start a new team focused on using Claude to accelerate pre-training research itself. Bringing Karpathy in to lead a team that uses Claude to make Claude better is the clearest public commitment yet that frontier labs see AI-accelerating-AI as a competitive lever. For working engineers, this cuts both ways. If the bet works, the capability curve steepens since every Claude release becomes leverage for the next one, and the timeline for AI handling more of the work shortens. If it doesn't work, it confirms that scaling has limits and the frontier is harder than the labs have been suggesting.