AI Working Principles (Karpathy-derived)

Five working principles for an AI coding agent, shared via Reddit as an extension of
Andrej Karpathy’s original 4-rule CLAUDE.md: ask rather than assume; match solution
complexity to the problem; stay in scope but surface issues found in passing; flag
uncertainty and prefer small low-risk experiments over false confidence; proactively
suggest better long-term approaches, not just tactical fixes.