Take some job you do a lot and write down the steps. Number them, 1, 2, 3 etc. Then step back and look at the list and think. The same way you might think about an interesting crossword puzzle, or a bridge game. A SimCity. Your goal is to eliminate steps.

I find that if there's an underlying simplicity to the problem, if it's possible to simplify, you'll figure it out.

Now, of course sometimes to simplify you have to cut off options. Another way of thinking about this is you're paving a cowpath. Look to see what people do the most, where the path is most worn, where the tracks are deepest -- where the taverns and hotels went up. These are the places where it pays to smooth things out.

A great example of this was Think C in the early days of the Mac. Think C had a shorter learning curve than a powerful command-line tool with a shell language. That was the downside. On the upside -- you could start having fun sooner. And the downside wasn't all that bad, because patterns had emerged, they were basically codifying and supporting processes Mac programmers were already using.

I think we're in the process of creating the cowpaths now, in the emerging environment of static apps and server-side JavaScript. It's an interesting time to think about it, because development models are simplifying on their own, and with a little willfulness we could get there faster.

Here's an A vs B to consider. I have a Node app to deploy. You could create an EC2 instance, or you could run on Heroku or Nodejitsu. In one case you get the full flexibility of a virtual PC. All the options. But it's still interesting to be able to drop a JS file in a folder, type a few commands at the shell, and have it run in its own process on a machine you don't have to even think about.

Ultimately, I'd like to be able to deploy by choosing the Save command from a File menu in my editor. That would be the easiest. A 1-step deploy.