This is already getting a ton of play around the blogosphere. Let me add to it and introduce you to mapnificent, which maps pedestrian/transit sheds based on location and amount of time to determine the effective "radius" you can travel in that period of time.
Even though Atlantic Cities is way ahead of me (and apparently with way more time) in mapping cities at even scaled comparisons I put three together (so far), though many foreign cities are missing and as yet unavailable.
I mapped three cities from their "heart" or primary crossroads, NYC from Times Square, London from Picadilly Circus, and Dallas from Pegasus Plaza.
As you can see London has by far the largest area that it covers. NYC is surprisingly small and you can see what a barrier large water bodies can be.
A couple of quick thoughts and limitations:
1. Like any data mapping device, this is entirely quantitative. You can't really derive any information based on quality of place from it. It is merely a measure of true universal accessibility. And accessibility without the expense of the car and the pressures and costs on infrastructure of everybody being in the car instills opportunity and a certain amount of independence. It's a way for anybody to reach a larger market for their ideas, their business, etc.
2. This can't take into account headways (yet?), the frequency of transit service or the time between buses and trains. Otherwise, it would be tethered to the moment you input the data (or as with google maps direction planner, any time you wish to begin your trip), which may very well be a next step for the creator. For instance, it doesn't give London the credit it deserves for frequency of the tube (though it doesn't run past midnight. Frown.) Doing so would add a pretty incredible amount of noise into the data, but it would be helpful. When I was in London last month it was truly astonishing that every time we stepped to a train platform a train was either waiting or pulling up. And we rode it quite often. This means when we're running 15 minute headways or greater, that adds that much more time to the trip, thus narrowing the true radius. Therefore, a low density place like Dallas which can't support more frequent headways will be overrated by this site.
3. Another next step will be to tether this to regional rail (HSR) data and air passenger routes as well. This is my dream. Can we measure how far you can get anywhere in the world from any one place in 1-hour? 2-hours? 4-hours, etc.? (and factoring in headways). Doing so would allow us to decipher the MOST connected (and least) places in the world. By altering the time gradient to 10-minutes, 30-minutes, and a few hundred minutes, would allow us to determine the local and global connectivity factors and thereafter begin to draw some correlations between global+local connectivity and land value, density. Tapping into open source mapping data is really expanding boundaries of thought and understanding. Yet the most exciting part is this is just the baby steps.