Mathematica is freely available on the Raspberry Pi[1] and most universities have site-wide licenses. The "Home & Hobby" licenses aren't even that expensive, only $195/year for a subscription or $390 for a perpetual license (and only $175 for renewals)[2]. And honestly, for those who are interested in tinkering but can't afford those prices, cracked copies are neither hard to find nor hard to install.
Personally I am quite fond of Mathematica (well, the "Wolfram Language") and I'm happy to pay for the hobby license price. Not only do I think I'm getting my money's worth, but I think supporting mathematical software is a "good cause" to put money behind. Moreover, I've never understood why people like amateur photographers will often spend more on tools like Adobe CC than many programmers (amateur or otherwise) spend on their tools period. Or why someone would spend $20-40+/month on various subscription services only to balk at $200-$400 licensing fees. Then again, I spend more time in Mathematica than I do in almost any other program installed on my computer.
However there is still an important place for (F)OSS mathematical software. As comprehensive as Mathematica usually is, it still has some significant shortcomings in advanced mathematics. In particular, I have a hard time believing it will ever fulfill those more "niche" areas of mathematics for two reasons. First and foremost, the ROI drops off a cliff for more advanced and/or esoteric fields. Second of all, the Wolfram Language already has 6000+ built in functions, so adding hundreds more to comprehensively support of something like Group Theory just doesn't make sense. Sure, they could be supported via packages, but that comes at the cost of performance (no first class support in the kernel) and usability (users have to go out of their way to use it).
Therefore, (F)OSS software like GAP, M2, and PARI/GP serve an important role in "rounding out" the gaps in the wolfram language. In my case, I contribute to FOSS projects an equal amount to whatever I spend on my Mathematica license. Or when monetary contributions to a project are not so straight forward, I try and contribute my time and skills improving them.
To be frank, I don't care much for projects trying to replicate Mathematica's functionality. Of course people are still going to develop and improve them, which at the very least puts some pressure on Wolfram Research to constantly improve basic functionality, but it will probably take one or two decades for said projects to replicate what Mathematica/WL is today.
Stephen Wolfram seems to get a lot of hate on HN, but for doing experimental math, puzzle solving, quick data viz, and much more, Wolfram is a deep and lovely language. The integrated notebook, mouse hover docs, and (shockingly) single massive namespace with thousands of functions somehow add up to a simple and productive experience unlike any other I've used. And I am not generally a fan of "thick" IDEs.
Adding the notebook GUI to a CAS (with a flexible programming language like Lisp or the Mathematica language) was a huge innovation, and we can probably thank Mathematica (along with MathCAD and Maple) for providing the inspiration for other notebook systems such as Jupyter.
The GUI is a big part of what made Mathematica better (for many users) than Macsyma. But it wasn't just the GUI - it was also the document-oriented design where you could save the graphs and computational results, as well as the code to generate them, in a single document.
Python had a perfectly functional REPL, but Jupyter notebooks were still a huge advancement.
"I'm happy to pay for the hobby license price." ..." Then again, I spend more time in Mathematica than I do in almost any other program installed on my computer."
Personally I am quite fond of Mathematica (well, the "Wolfram Language") and I'm happy to pay for the hobby license price. Not only do I think I'm getting my money's worth, but I think supporting mathematical software is a "good cause" to put money behind. Moreover, I've never understood why people like amateur photographers will often spend more on tools like Adobe CC than many programmers (amateur or otherwise) spend on their tools period. Or why someone would spend $20-40+/month on various subscription services only to balk at $200-$400 licensing fees. Then again, I spend more time in Mathematica than I do in almost any other program installed on my computer.
However there is still an important place for (F)OSS mathematical software. As comprehensive as Mathematica usually is, it still has some significant shortcomings in advanced mathematics. In particular, I have a hard time believing it will ever fulfill those more "niche" areas of mathematics for two reasons. First and foremost, the ROI drops off a cliff for more advanced and/or esoteric fields. Second of all, the Wolfram Language already has 6000+ built in functions, so adding hundreds more to comprehensively support of something like Group Theory just doesn't make sense. Sure, they could be supported via packages, but that comes at the cost of performance (no first class support in the kernel) and usability (users have to go out of their way to use it).
Therefore, (F)OSS software like GAP, M2, and PARI/GP serve an important role in "rounding out" the gaps in the wolfram language. In my case, I contribute to FOSS projects an equal amount to whatever I spend on my Mathematica license. Or when monetary contributions to a project are not so straight forward, I try and contribute my time and skills improving them.
To be frank, I don't care much for projects trying to replicate Mathematica's functionality. Of course people are still going to develop and improve them, which at the very least puts some pressure on Wolfram Research to constantly improve basic functionality, but it will probably take one or two decades for said projects to replicate what Mathematica/WL is today.
[1]: https://www.wolfram.com/raspberry-pi/ [2]: https://www.wolfram.com/mathematica/pricing/home-hobby/