@luis_in_brief the tool's benefits are immediately obvious and overstated, and the costs are insidious, hidden by a spectrum of dishonesty from sample bias to outright fraud, extremely difficult to measure, often externalized, and systemic. my consistent position is that usage tends to be _net_ negative. there are places where it can locally trend positive (assuming you ignore training costs), but there are a lot more places where it *appears* to trend positive