Yes, but then you are also dead wrong.
Notch writes terrible code.
Personally I maintained various codebases (not written by me) and some of them were really terrible, utter mess, while some others were pretty nice to work with, even if I had to refactor sometimes. If programmers thought about code design at least for a while before writing it, and then the code is well-organized (= good separation of concerns, no global state, no copy-paste, complex parts isolated from the simple parts) - it doesn't matter that much that they didn't think of a new feature you're now adding to it.
There were a few articles written up about a year ago [1] from interviews with Notch. He does stuff like direct variable access instead of using setters/getters because it "gets in the way" which when it comes to encapsulation and separation of concern, etc. it throws off some red flags.
I've looked at some of his (at least I believe they were his) projects when attempting to learn a game library (libGDX) and was kinda grossed out by the code. You can say that it's because it was from a Ludum Dare perhaps, but you have to think that he kinda does this all the time. So what would be different from his production stuff.
Pretty much don't look at him as a programming idol, he's simply too cowboy for that. That's fine, cowboys have built tons of stuff, but if you are trying to learn something you may want to avoid their stuff cause it's probably running against the grain in a bad way.
[1] The Article -- http://notch.tumblr.com/post/15782716917/coding-skill-and-th...
In my opinion it genuinely does get in the way. I want to add a member variable to an object - in many languages (JS, C#, Python for example) you can just declare it and move along, knowing that you can switch it to a property later if you need special handling.
In Java, however, for future maintenance I have to write a `public int getPaula()`, `public void setPaula(int brillant)`. This not only violates YAGNI, it slows things down in the physical "I have to type more" sense.
Thankfully C# makes declaring getters and setters easy from the start, so it's not an issue to use them:
public int SomeField { get; private set; }
That gives you a public property called SomeField which can be retrieved, but which can only be set privately by the class. I'd hate having to write GetBlah and SetBlah; it's damn ugly.
Its a single-app code-base. Its not an API anyone consumes. Using get/set boilerplate really is just getting in the way.
That's the area where architecture is driven by philosophy. It's whether or not you _choose_ to study the problem, search for best practices, and work out contingencies during the design phase, or go forward with a simplistic, unknown solution and study it by driving trucks over it[0]. With most exploratory programming - of which games are a prime example - you want to drive trucks because your design is likely to be reworked many times regardless.
As well, independent creation not driven by clients and markets has to be done entirely selfishly. When taking this to heart, one naturally moves towards an exploratory style, because selfish, truck-driving code will satisfy your urges more quickly.
According to some links posted elsewhere in this discussion, Notch admits himself that he is not the world's best programmer, but can we honestly say his code is bad enough that we can consider him one of the world's worst programmers?
Or are you really trying to say his code is average to good, with some mistakes made along the way?