Editor's Synopsis
A Twitter user summarizes a recent insight from Andrej Karpathy about a shift in how he uses AI. Rather than primarily using AI for code generation, Karpathy is now spending more time leveraging LLMs to build personal knowledge bases. His workflow involves dumping raw sources — articles, papers, repos, datasets, and images — into a folder, then having an LLM organize and structure the material.
This reflects a broader trend where AI power users are moving beyond code assistance toward using LLMs as knowledge management and curation tools. The approach treats AI not just as a productivity shortcut for development, but as a way to synthesize and make sense of large volumes of information across diverse formats.