Jobs today ask us to “wear more hats” than ever, drawing academic attention to the emerging trend of skill diversification in the labor market. Most research on changing job skills focuses on high-paying occupations and treats skills as isolated boxes to check, ignoring how skills connect to one another and ignoring the possible skill diversification in low-wage work entirely. This talk introduces Skill Embedding, a new framework that maps skills onto a shared space to reveal how they relate, and applies it to millions of job postings and representative survey data. The results show that skill diversity within jobs has surged three times faster in lower-status service jobs than in higher-status management and professional jobs, and this shift is driven largely by workers in the lowest-status roles, who are diversifying their skills fastest but earning the smallest rewards for it. These findings reveal overlooked skill dynamics in low-wage work and demonstrate the value of a system-wide, distributional approach to analyzing skill relationships.