Wolfram Language and Mathematica version 15
writings.stephenwolfram.com
writings.stephenwolfram.com
The problem though is that Wolfram is a walled garden. When you think about integrating it in an enterprise environment, you get hit by such high costs, it stops making sense. Imagine if they open sourced it, I feel like their products have so much utility, buried deep down Wolfram ecosystem and conventions.
Many years back while in grad school I could not reproduce a result from a paper. Thankfully they had provided the data as public but not the code. I emailed the authors and got some matlab code back. My university didn't have a matlab subscription. Octave saved me there since the syntax is similar.
But with something like mathematica and the price of it you will never be able to have a wide verification of the result if the software is not free.
Also, a lot of things in industry gain traction first in academia (especially math tools). So unless academic traction is dealt with mathematica's headway in industry will remain limited. They are still a profitable company. So I'm guessing there are deep pocketed clients who purchase the tooling.
Even the Python GPU JITs only started to be invested seriously by the big three last year.
It will take a while.
This is not strictly true. The Wolfram Language/Engine is free. You can run the code in the terminal or after connecting the kernel to Jupyter Notebook. The frontend Mathematica is not free.
https://writings.stephenwolfram.com/2019/05/launching-today-...
https://github.com/WolframResearch/WolframLanguageForJupyter
Reimplementation in Rust: https://github.com/ad-si/Woxi
WLJS Notebook: https://wljs.io
VS Code extension: https://github.com/vanbaalon/wolfbook
It's so disappointing to see CLAUDE.md in projects like these. Basically rules it out for serious use.
Spot checking, I don't see any issues.
e.g. https://github.com/ad-si/Woxi/blob/main/tests/list_tests.rs
#[test]
fn first() {
assert_eq!(interpret("First[{1, 2, 3}]").unwrap(), "1");
assert_eq!(interpret("First[{a, b, c}]").unwrap(), "a");
assert_eq!(interpret("First[{True, False, False}]").unwrap(), "True");
}Contrast with Julia where it can be a regular Julia library,
I don't know that you can match something speedwise like a JIT or Expression Templates in rust though without using something like Enzyme.
A very young project written by AI means you haven't reviewed the code and nobody has used it in anger. It might work perfectly, but my experience of AI so far says that it won't.
For example, when digging into GNU Octave you will find many of its libraries were built on peer reviewed legacy code provably reproducible with prior aerospace published works.
The problem with closed source academic programs isn't features or even quality, but rather one of traceable Metrology and scientific rigor. =3
https://www.statista.com/chart/4111/do-europeans-wash-their-...
[1] a personal perpetual license is only $400.
There is a combined license for Mathematica + System Modeler, but it's "Contact us for pricing". Mind you, that's still on the Hobbyist tier. You cannot use its output for anything commercial.
Contrast with Julia's MTK/Dyad that are free for non-commercial use.
https://www.thelinuxvault.net/blog/how-to-run-the-raspberry-...
It doesn't cover the full standard library of Mathematica but the syntax is very similar and a lot of functionality is there.
Its standard library is almost impossible to reproduce in its enterity
If those libraries were like regular code that got published to Github or something like that.. like pypi or npm or crates.io or whatever. And if mathematica had a lean standard library. It would be very feasible to implement a clone that's basically compatible
I mean. Depending on just wolfram rather than random open source contributors has benefits, for example it's more resistant to supply chain attacks. Indeed the npm model is not good. But, it is open, and that's what enabled for example deno and bun to have some compatibility with node
Like the following dudes who are doing this, but to a project that is already open source (git):
https://news.ycombinator.com/item?id=48468904
Except that Wolfram would of course sue and we might as well see what the courts has to say about this topic
This sounds like your average functional programming language. The Scicloj community is the first thing to come to mind (but I assume they don't do symbolic algebra/calculus like Mathematica does), but I don't know what you're specifically missing.
But a walled garden that costs $400 for personal use (we're ignoring yearly licensing, because f that noise) is utter nonsense, and the clearest sign you have no idea how to sell and then upsell products to users over the course of several years.
A quote from the linked article: " ...year after year building an ever taller tower of ideas and technology ..."
That's an accurate description of the Wolfram empire -- every year it becomes a more expensive, less accessible, vertical tower. Meanwhile, people intent on disseminating useful knowledge do so by growing horizontally -- Python, Linux, many others, all open-source.
Historical figures would be astonished at what Wolfram is trying to do -- they would say, "Wait ... you can't patent mathematics!" No, but you can try.
That's surprising considering how good their documentation is. A tool using LLM should have no problem with that. WolframLanguage is almost ideal for an LLM actually.
I don't really like English from a linguistic point of view (as a non-native speaker). It's a hodgepodge of other languages and has so many exceptions, it's not very elegant. But it's so ubiquitous and useful that one basically has to know English today.
On the other hand, Latin is beautiful and pure. There's more rules, but very few exceptions. But unless you study catholic theology or something along those lines, it's basically useless.
Which one maps to Wolfram Language and which one to Python is probably obviously.
There are irregular verbs, sometimes with complete suppletive replacement of principal parts by what used to be other verbs (e.g. sum, esse, fui, futurus; fero, ferre, tuli, latum). There are verbs that use passive forms with active meaning (deponents) or perfect forms with present meaning (defectives).
There are arguably completely missing forms in the verbal inflection system (the Romans knew that some forms plausibly "should" exist, especially based on a Greek grammatical model, but simply didn't have them!).
There is sometimes unpredictability in which noun case should be used with a particular verb.
The noun declensions are apparently based on two different sets of Indo-European noun inflection paradigms, so nouns with similar nominative forms can end up being declined very differently.
There are ambiguities where different noun forms coincide, which can even create parsing ambiguities in literature (like confusion between ablatives and datives, many of which look identical).
The extent to which the perfect stem of a verb can be predicted from the present is limited, as sometimes stem reduplication is used, but sometimes just suffixation of something like -vi.
There are loanwords, even classically, from Etruscan, Greek, and to a lesser extent other Mediterranean languages (just thinking of that "hodgepodge" issue).
The meanings of purpose clauses with the verbs of fearing are arguably backwards from the English point of view (although I think the Latin version does make plenty of sense).
Native and nonnative speakers couldn't easily agree in antiquity about whether vowel length should be contrastive and (I think) whether consonant aspiration was phonemic. I guess the native speakers' opinion should matter more, except there promptly became such huge numbers of non-native speakers that they started to have a really humongous influence on the language.
There are spelling changes even within the classical period, so there isn't quite one single classical Latin orthography.
I guess there are many fewer irregular verbs overall compared to Germanic languages (which historically have had up to hundreds of at least partly irregular verbs). But if we want to count unpredictability of Latin perfect stems (which is somewhat akin to the main source of irregularity in the Germanic verbs: stem changes) as a kind of irregularity, Latin will also have quite a lot of these.
1)the locative vs the ablative, and the locative only existing for a few words
2)the irregular verbs such as sum, eo etc, irregular nouns such as deus, aqua etc, and there’s a bunch of irregular like adjectives and stuff that I don’t remember
3)indeclinable nouns that just don’t decline at all and are the same in all cases. I think the word for “morning” is like this but it’s been a very long time. There are a few words that work this way anyway.
4) Words like “castrum” which just mean something totally different in the plural to the singular. “Castrum” means a fort, but the plural “castra” doesn’t mean many forts, it means a (singular) military camp.
5) Words like “Saturnalia” (festivals of Saturn) which only exist in the plural. As far as I know you can’t say one festival of Saturn in latin.
For (5), these are called pluralia tantum (singular "plurale tantum").
https://en.wikipedia.org/wiki/Plurale_tantum
I gave some examples of Latin irregularities elsewhere in the thread, and I like your examples too!
I've programmed quite a bit with both Python and Mathematica, and I've read through your comment a few times, but I still can't figure out which is which. Both languages are hodgepodges of other languages with lots of special cases (which I would consider to be a good thing since it gives you so much flexibility).
People joining my company from academia usually know Mathematica along with Python or R.
When we tell them we don’t use Mathematica they are sometimes initially concerned. They are typically quite opinionated and I have yet to hear an employee complain about no longer having access to Mathematica. Or SPSS, SAS, or MiniTab for that matter.
Claude Caude is much better at Mathematica than Wolfram's own AI assistant. I think they flat-out acknowledge the very limited abilities of Mathematica's AI assistant in this version 15 announcement.
The Wolfram AI assistant is so bad I unsubscribed from it. By the sounds of it, a basic AI assistant is offered included with subscriptions now. I feel it's borderline criminal they were charging for their hallucinatory AI assistant in the past.
But that's fine. Mathematica client supports openrouter as LLM provider anyway so we can use whatever we want.
SJ recommended some of the UI bits of the notebook. Particularly the separators between cells.
Wolfram had come up with some normal techie names "computron", "math-o-matic" or whatever. SJ said No those suck, use something simple like Mathematica.
Of course there is a whole constellation of more specialized things in certain fields, that has come a long way in the last 15 years. So people needing things like that no longer kludge things together in Mathematica.
https://www.12000.org/my_notes/CAS_integration_tests/reports...
Reimplementation in Rust: https://github.com/ad-si/Woxi
WLJS Notebook: https://wljs.io
VS Code extension: https://github.com/vanbaalon/wolfbook
https://www.12000.org/my_notes/CAS_integration_tests/reports...
Note that alternative open source solvers like Fricas fail 10x the integrals in that corpus.
But indefinite integration is just a small aspect of CAS capabilities. What about integration over a line or surface, definite integration and dealing with singularities, differential equations, solving equations under assumptions, simplifying equations.
https://www.12000.org/my_notes/CAS_ode_tests/index.htm
He does a few other side-by-side comparisons but doesn't include open source engines in them.
I would not call him a crank, he's just an "independent physicist". He is a very successful and wealthy businessman thanks to Mathematica, with $100Ms of wealth presumably, and he choses to do physics in his own way, pursuing subjects that he finds interesting, in ways he finds interesting. And he writes about his work and himself in grandiose ways, usually comparing himself to Newton and Einstein.
Nothing major has come out of his research, other then one of his co-workers proving that one of the simple CAs is Turing complete.
Most academic physicists ignore him, but that's fine. Personally, I think we need more people like Wolfram who are doing totally independent research, with their own funds. Statistically, something unexpectedly good could come out of it!
His latest research subject is Ruliology:
https://writings.stephenwolfram.com/2026/01/what-is-ruliolog...
> And he writes about his work and himself in grandiose ways, usually comparing himself to Newton and Einstein ... Nothing major has come out of his research
okay
Wolfram does display a similar ego to cranks. He tends to place his cellular automata at the same tier of importance as general relativity or wave-particle duality. But unlike cranks, he doesn't say that Einstein, Feynman and other greats are wrong, he doesn't "prove" his theories by math-looking babble that is definitely not math. He loves cellular automata and so much of what he writes is more like philosophy of science than science. He will model certain processes as an automaton and then he jumps from that to every natural process being a cellular automaton.
Eccentric, definitely, and most of his physics research isn't accepted but he's several orders above the level of crank.
Except stealing credit for work done by his employees. Read about his lawsuit against Matthew Cook.
I find greater fault in the pushing of his "Intellectual Dark Web" pseudo-cult. So I would agree he's actually a contemptible character but for slightly different reasons.
what field?
Unfortunately they are extremely surface level in this release. It looks like there isn't even any ability to load/export MusicXML, kind of weird low level primitives and zero interesting higher level functions. Hopefully they keep iterating on it but I don't think it'll be useful for my workflow right now.
I would rather live in a world where my life is surrounded by mystery than live in a world so small that my mind could comprehend it. - Harry Emerson FosdickThis is because clicking the "granulated sugar" selection just appends an override string to the URL, turning the query into a very generic one for "weight of granulated sugar" using API style query semantics. The original input text is still rendered in the box because it was still left at the start of the query string and the override isn't in input text form to replace it. One can see the same result by directly putting "weight of granulated sugar" in the input box, where 1 tsp is just the default quantity used when generating examples with unspecified amounts for most ingredients.
If I were to guess, this was several independently made UX/UI decisions ("leave the originally structured input text in the box as modifiers are clicked", "have default quantities of a material to still generate an example result when a generic query is made", "suggest different types of material to work with", and "a dropdown to select the material should be a modifier") which made sense for isolated components but, later, the components were combined in this flow and nobody tested to see the aggregate behavior no longer makes any sense. Ironically, the issue would likely not exist if modifying the query for the new type was just given to LLM to figure out dynamically in the back end instead of trying to explicitly code the behaviors in the front end.
Wolfram product front ends have historically been broken in little wonky ways like this. Hell, they had a completely half-assed "dark mode" in Mathematica's GUI client until 2025 and just now have added actual visual themes to notebooks in 2026 - the attention tends to be given on the parts beneath the covers.