A randomly-round milestone (which does not include 424 answers posted under another account) reminded me of how my Stack Overflow activity shifted over the years. In rough chronological order, it went through the following.
Matlab and Scilab is where I started, being familiar with both from some Numerical Methods courses. Downsides: most Matlab questions are either answered quickly, or are unanswerable. Scilab is a much smaller niche, with almost no competition from other answerers. But a low-activity tag also means virtually no votes. Eventually I got bored of both and do not watch these tags anymore. Sorry, Scilab – I like your improvements on the weird Matlab syntax, but the emergence of Python’s scientific stack leaves you with little potential userbase.
Google Sheets is a topic awkwardly split between Stack Overflow and Web Applications. Is using spreadsheet functions programming? Well, is writing regular expressions programming? Is writing SQL queries programming? Google Sheets functions include both regex support and SQL-like language, although the latter is seriously limited by the lack of joins. It can be fun to find workarounds for various limitations of spreadsheet functions (example). But it does not last forever; I have long abandoned both the Web Applications site and the Google Sheets tag on Stack Overflow.
Google Apps Script is the tool I use daily. It processes bundles of spreadsheets, it scrapes information from various places and serves up static webpages and simple web applications for my own use. For example, when I’m in the mood to use some moderation tools, it identifies the posts that need to be disposed of.
SymPy is a library I almost never use for my own work, but I like this mix of pure math and pure Python. Not to mention plentiful bugs or deficiencies of implementation, which create fertile ground for new tricky questions. On multiple occasions, answering a SymPy question led to diving into the source, and possibly adding an issue, sometimes with a pull request. This is the tag I watch most closely now.
NumPy and SciPy, of course. I like interpolation questions the most, followed by other typical numerical tasks: integration, optimization, root finding, numerical linear algebra. Tricky situations arise all the time, and the sparsity of SciPy documentation in niche areas is remarkable. Shockingly, one still needs to know math to do math with a computer.
Although I dabbled in both pandas and scikit-learn, I left both tags pretty quickly: pandas has no shortage of extremely active answerers, and the scikit-learn tag is depressingly full of “I got a machine learning problem and installed Anaconda, what do I do next”.
For a very long time now, Stack Overflow has been headed in two different directions…
The first direction involves building an increasingly-detailed set of Q&A on broad, popular topics. This is the easiest one to observe: if you’re learning, say, C# then you’ll find a very broad and deep set of information here, with more and more corner-cases being filled in hourly. The single biggest obstacle here is noise: with almost every possible topic covered in multiple questions expressed in multiple ways from multiple perspectives, finding a question that seems to match the problem you’re having is easy – but finding the question that has an applicable answer can be a slog.
The second direction involves covering an ever-more-vast set of topics. Everything from languages and platforms with only a few (initial) users, to specialized libraries and tools nominally under the umbrella of a more well-known language or platform. The big problem here is (and has always been) that these topics are nearly impossible to moderate effectively; they may have only a tiny handful of active users, they tend to not attract much voting, and the bulk of the moderation tooling (and elected moderator team) is skewed toward serving #1.
Solving the problems of #1 is… An almost insurmountable challenge. We tried hard to attack it head on, and eventually came to the conclusion that we could probably throw a billion dollars in dev time at it and still only maybe succeed; the root causes are simply a lot bigger and broader than the little Q&A site they affect. That doesn’t mean we shouldn’t try to mitigate those effects, but it very much means we should do so while compulsively reciting the serenity prayer.
Because of this… Or perhaps in spite of it… I’ve come to believe that direction #2 is the future of Stack Overflow, the area that both askers and answerers will find most rewarding in years to come. Specialization may not appeal to Lazarus Long, but as our field matures it’s an increasingly-necessary option – and one we have a real opportunity to serve better here on Stack Overflow.