Your painting analogy is really solid. There is a remarkable range in the way that software needs to scales and that captures it.

    Doesn't it depend upon WHAT it is that is produced? 

A productivity issue that doesn't get discussed often is the bridge between software and the people who work with it. For example - developers often draw the line at "works on my machine" or "I can deploy it to production" or "I can work out bugs". But these things don't take into account the scenarios of customers, deployers and support operators - documentation, logging, troubleshooting and the amount of time spent by those customers and anciliary staff doing those things. Effectiveness at these things is generally far more important than whether a programmer can write something in one day or five, but the typical developer is a blinkered creatures.