hhhmm...interesting. 
I guess we have to wait and see if someone wins. 

Off the bat, you have a 1/1000 chance of winning unassisted.

With a program you could have it post 30ish times, and get the difference between the submission time and server time to find out network latency. Get the 95% confidence interval of all those trial latencies and start posting that much time *before* the beginning of the next second. Should take about 20 tries after that depending on the variance. 

I imagine even with a program it's going to be hard due to network latency etc.

But I'm going to write a script to verify timestamps!

First Comment Wins (Almost)

what? This is going to be impossible...
Unless someone writes a program for it..