I study how LLMs think internally: features, circuits, and residual streams.
Treat this blog like lab notes shared out loud.
