I first heard it in the late '80s, in the context of grad school admissions, but perhaps Ousterhout was the first to port it from math to computing.

For a real treat, try replacing a curve that is fundamentally linear (say, optimizing your salary as an employee) with one that is fundamentally exponential but with a much smaller constant factor (say, founding a startup). Not only will you be below the Y-intercept for a considerable period of time, but you will also have a slope that is indistinguishable from zero.

This is probably the biggest killer of startups: it doesn't feel like you're making any progress at all, until you have a sufficiently deep picture of your marketplace to know exactly where to hit it that your competitors aren't, and a sufficiently solid product that your customers realize this.