I would have liked to see some sort of continuous integration going on. The assets are the winning clip, athlete (name, gender, sport), winning result, and whatever other metadata they need. Once all those are fed in, a build process is automatically triggered which constructs the completed commercial by integrating the clip into an appropriate base commercial (overlaid green screen, audio track), outputting the result on a whiteboard by displaying the first part of the time off-camera using a hand-drawn font (for the "4:05." part in the commercial) and then animating the hand to draw the remaining digits.

Once all is rendered, it triggers an approval process by the agency and then by AT&T/NBC and, 5 minutes later, the commercial goes live. That's how I'd want to do it. Perhaps there is also Olympic committee approval, FCC involvement... who knows what other regulations they need to jump through.

Yeah.. not exactly hackerish..

They did the same shot 100's of times and just picked the one off the shelf that matched the swimmers time.

I expected them to use a hand-model for watching "someone with a sleeve write on a whiteboard".

I have seen adverts for live odds* which, according to the smallprint are up-to-date within 30 seconds.

*not Olympics