Looks great! I agree that there should be a blurb to the effect of "predicting hotness while it's still lukewarm," or something less lame but more informative. Do you use a form of Poisson regression, by any chance?
I do like the mix of stories. To be honest, I probably won't use it tomorrow. But you know what? I would probably use it tomorrow if you juxtaposed (a) the time when you predicted it would be hot, with (b) the time when it became hot by some other metric. Obviously the newest stories wouldn't have that yet, but the ones that are ~30min old just might. In other words, if I felt like I was getting in on something special, I probably would use it.
Yes, totally agree. If you indicated how much later it appeared on websites such as Techmeme, Google News, maybe HN, I would totally come back to this. Also might be nice for comments sometime down the line?
Do you like the mix of stories? Do you think you will use it tomorrow?