That's simply not true. A lot of these projects provide software with excellent capabilities. The problem is, after someone writes a good algorithm the project

That's simply not true. A lot of these projects provide software with excellent capabilities. The problem is, after someone writes a good algorithm the project is not done. Producing good end user documentation, a user interface that makes sense, or choosing UI paradigms that enable casual users to be productive - those are tasks which can easily take up the majority of a developer's time. A good case could be made that many open source developers don't even recognize these areas as goals or see how their project could possibly be structured any differently. This results in a lot of wasted time for the users, use case failures, a lot of duplicated effort across inaccessible projects, and the impression that these programs are not efficient. However, by far not all GNU projects look like this - a ton of them could be considered "good as is" even by commercial standards. For example, Octave just works as it should, is well documented, and doesn't fight its users all the time. Even Gimp recognized usability was important, they just got it completely and mind-bogglingly wrong.