Or invent a programming language that gives a lot of the guarantees and robustness of a hard-ass fp lanuguage but is as easy to pick up as Java or Ruby.
Such a broad topic. You should at least hint at some of your interests.
Ideally, do something related to your true interests in software engineering. 6 months is not a long time. You definitely want a deliverable code product, even if just a minimum-viable implementation.
If you are having trouble remembering, go back through your bookmarks/upvoted HN posts (if this a throwaway)/read laters.
Suggestions:
Take what ImTalking said and apply it. Find an industry that could benefit from software improvements/modeling/automation. Find the pain points. https://news.ycombinator.com/item?id=11444451
Combine data visualization with something of interest to the HN community that could go "viral" on HN once you publish it.
Write an interesting niche chat bot (this is a really emerging field). Perhaps with machine learning.
Write something that your future employers might use as a hireability signal
Start a blog. Post it on HN. Document your process of researching and writing the thesis. Show us what it is like to be researching something in depth. I think many at HN (even those who openly eschew the idea of going back to school - or even going to college in the first place) would find it interesting.
Thanks for the advice. Would love to post about it - hadn't really thought that it could be of interest to anyone :-)
Mostly my interests are within cloud computing, machine learning and web development, so more or less mainstream fields. E.g. a topic could be to utilize some cloud computing patterns or machine learning in a given setting. I just lack inspiration for the setting.
I didn't continue with education after my degree mainly because I wasn't sure of what I actually wanted. I'm interested to hear what others did for their Masters/PhDs and what ideas people have.
If I had six months to research something, I'd work on temporal databases, starting with the book Developing Time-Oriented Database Applications in SQL. I'd look at the history of attempts to bring this into the SQL standard, including the arguments between Snodgrass and Date, and what was eventually put in SQL2011. Then I'd see how much of this Postgres supports, and possibly take a shot and filling in some gaps. Or I'd try to adapt a popular ORM to work with temporal tables.
But that is me. Pick something that interests you!
By the way, for six months, you had better pick something very focused. I would try to err on the side of modest. If all you have right now is a broad topic, see if you can pull at one loose thread---some annoyance or nagging question---and see where it takes you. Probably you will find hundreds of articles to read, and more complexity than you expected. Grad school is a rare chance to "go down the rabbit hole", but you should expect that your six months will fly by. Research is like reading Wikipedia: it constantly branches and takes you into new things you "need" to know.
A problem in my field, scientific computing and HPC, suffers from a big problem involving the reproducibility of experiments. Some tools like Jupyter help, but that is only useful in a limited capacity.
If you have any interest in helping to improve how mankind organizes its knowledge (as a whole or at any level), I'd love participation here (or just your exploring it in an academic way): http://onemodel.org (AGPL, a new approach I call "atomic knowledge" because of the internal structure, which I discuss if you explore the web site a bit, but it could be described better than it is).
Code sniffers and linters and style checkers are tools that detect that a bad smell is present.
An architectural bad smell is violation of the architecture. IE the view modifies data, the controller is too fat, etc.
Hence it is required to have a Code sniffer for architectural bad smells.
However considering that there are so many architectural patterns, so many frameworks that implement them it is herculean task.
So could it be possible to use Machine Learning to detect them hence the thesis:
"Automatic detection of architectural bad smells with Neural Networks"