Not your Weights, Not your Brain (mercurialsolo.github.io)

In a recent discussion, AI expert Andrej Karpathy emphasized the transformative potential of what he terms an "exocortex," where our cognitive capabilities are augmented by AI and computing technologies. He highlights the significance of ownership over decision-making processes—the need to retain control over the judgment layers generated from our interactions with AI, rather than allowing this valuable knowledge to be absorbed solely by external models. This shift from merely running AI locally to capturing and retaining personal judgment signals is key to maximizing the value derived from AI systems. Karpathy illustrates this with his tool, claudectl, which enables users to log and refine their decision-making processes with AI in real time. By continuously correcting and approving AI outputs, practitioners not only help improve the AI but also enhance their own cognitive frameworks. The captured judgment traces, rich in context, can inform future AI interactions and provide a competitive edge in professional environments as the focus shifts from reliance on basic AI capabilities to our own distilled skills and experiences. This personal management of AI-derived knowledge showcases how the future of work will be defined by the synergy between human expertise and AI assistance.