I can't say I blame him for putting up a fight, since it's his bread-and-butter and GPLing it would cut his financial knees out from under him, but it's not even a debate at this point whether or not GPL code is included in Thesis. This is a fight that he can't win.
(I'm not well enough versed in this particular debate to weigh in, but there is an answer to that question, in general, and it hinges on specific details of one's implementation of "creating something.")
In the end, it's the laws of the land that you're in that matters. As far as I know, I don't believe that these arguments on what is a derivative work are anything more than opinion. Settled cases are not equivalent to legal precedent.
The best thing that could happen for everyone involved is that it goes to trial and ends with a judgement, and then all this posturing and pontificating will hopefully come to a merciful end.
Pearson will most likely sue the first person who actually does this (and refuses the Cease-and-Desist that precedes the suit). The courts will then determine whether or not Thesis is indeed GPL, or whether the distribution was in violation of Pearson's intellectual property rights.
Until someone actually does this, the debate will continue - there's enough ambiguity that the issue won't be resolved until ruled upon by the courts. I really doubt Matt/Automattic will push the issue - there's a lot to lose and almost nothing to gain. (Matt said in a previous thread he thinks Thesis' code is "junk" and doesn't want it in core.)
Until Thesis is publicly distributed and legally challenged, the status quo will continue: Thesis will effectively be private code, and Pearson's position will prevail.
IANAL, but the following is my understanding based on the my lack of awareness of anyone licensing code as GPL (even the FSF) who has ever claimed that what you suggest is legal.
Thesis is currently under whatever proprietary license Pearson distributes it under. He probably is legally required to distribute it only under the GPL. There are two obvious ways he can deal with this: stop distributing it or distribute it under the GPL. In the event that he takes the latter path, then people will be able to legally copy/distribute/modify/fork/sell Thesis. However, for now, no one has any license to Thesis' code other than what Pearson has granted them.
The GPL can't force code to be released under the GPL. It can only forbid releasing it under other terms.
5. You are not required to accept this License, since
you have not signed it. However, nothing else grants
you permission to modify or distribute the Program or
its derivative works. These actions are prohibited by
law if you do not accept this License. **Therefore, by
modifying or distributing the Program (or any work based
on the Program), you indicate your acceptance of this
License to do so, and all its terms and conditions for
copying, distributing or modifying the Program or works
based on it.**
As the matter now stands, if it is true that they use GPLed code, that means that whoever distributes Thesis is guilty of copyright infringement, which means that the owner of the original GPLed code can sue Thesis for lots of money. But that does not mean that anyone can copy Thesis.
Put in other words, the viral effect of the GPL is not automatic. If you use GPLed code in your code it does not mean that your code is automatically GPLed. It means that you have a choice to (i) voluntarily GPL your code, (ii) get rid of the GPLed code from your code or (iii) be guilty of copyright infringement and face the consequences.
This is intended as an intellectual discussion of the matter and is not legal advice.
To put it as I understand it, using GPL code doesn't automatically make your code GPL; it places a legal requirement on you to either license your code as GPL, or to not distribute the source or product. If the author doesn't license his software, then nobody else can take and use it, but the author is in violation of copyright due to the unlicensed use of GPL code.
The virality of the GPL is a requirement of the license, not an automatic effect upon using GPL code.
Unfortunately, or fortunately, this doesn't seem to be the way suspected GPL violations are resolved. More frequently there is a community consensus and a public shaming of the responsible company (e.g. http://news.ycombinator.com/item?id=790316). The issue then becomes one of how much negative publicity the company is willing to take.
IOW although "GPL violations" are described in legal terms, could be legally tested, and refer to legal documents, the issue of illegality takes second place to a fuzzier issue of "fairness", which changes from community to community and project to project.