Today, one of the remaining appeals of Lisps (e.g., CL, Scheme) is that (aside perhaps from Clojure) they're not very employable, and so the communities and ecosystems are more old-school hardcore in sensibility.
In Python, I have to really get on a roll on a project in it, to where I can ignore most of the unfortunate aspects, that are apparent when having to Web search through writing and code that seems 99.99% by people who wouldn't be doing this if it didn't pay money. Many are skilled and professional and conscientious, despite the circumstances of it being a job, but that's not the norm.
In JavaScript universe, jeebus, you can't read about even the smallest things, without it cavalierly telling you to install a supply chain disaster from NPM.
In Rust, some of their linguistic thinking is appealing. They still have the positive-employability problem, :) but the difficulty has a gatekeeping effect that isn't entirely unwelcome (if you only want a job, just learn JavaScript or Python). I can foresee flexing some of my systems programming skills like I haven't been able to as much in Python, JavaScript, or Scheme, to build efficient and trustworthy software. But then, for example, the other day, I thought "I'll try this UI package, which is a wrapper around a platform Web rendering widget, since maybe they found a sweet spot..." But somehow, "hello, world" requires compiling 503 Rust crates, just for the wrapper alone, not counting the code for the non-Rust platform Web widget that the they wrap. We already know that the platform widget is going to be buggy with memory errors and design flaws, and we're going to be running bloated Web stack atop that widget, and with callouts to Rust code, and how many people in the world can even reason about debugging that (or auditing it!), just to display simple GUI widgets in a fashionable way.
Ironically, this author gets the relationship between lisp and writing totally wrong. Lisp may be much more artistic, but programming in Java, for instance, is much more akin to writing than programming in lisp is. Written languages have well established vocabulary and grammar, that cannot be changed or redefined by the writer. The author is completely correct that lisp is more of a "programming medium" than a "programming language", since the language itself can be molded and changed by the programmer in very self-expressive ways. However, he doesn't follow through with this observation to the obvious conclusion that this feature of lisp, as a medium, makes it fundamentally different from human language.
I love Richard Gabriel's writing. "The Structure of a Programming Language Revolution" (https://www.dreamsongs.com/Files/Incommensurability.pdf) is both insightful and beautiful.
When on Gabriel's site, take the opportunity also read about Cadillac, and Foundation for a C++ Environment.
How it became XEmacs, what the Cadillac protocols remind you of on a famous editor currently, the infrastructure for a image like development environment for C++, all around early 1990's.
This was a wonderful read.
It's a bit sad to see that the book he's referring to at the end is a book released in 2003, to which this is the foreword.
Lisp is an incredible experience to write and experiment in, unlike any other language I've ever worked in.
Still, I believe there are many people out in the field that are mostly about "just delivering the spec" or about making money; pragmatists who care not too much about the art behind the software, and I suspect that will always remain the case.
Craftmanship marrying art and mastery is a niche, but that doesn't mean it's not thriving in spaces for enthusiasts.
The author spends a lot time describing programming as a solitary activity (discovery, analysis) using like metaphors (poets, explorers, etc) and draws his conclusions thus, emphasizing the malleability of LISPs and Smalltalk as ideal for this task.
I think the author examining the fads of agile or XP would draw quite opposite conclusions if they observed programming as a social activity, building a shared knowledge and understanding that is constantly refined before it is "abandoned" as a piece of software.
lisp's power came from its default being openness. not just macros or homoiconicity, but how the system assumed you would reshape it. that expectation isn't present in modern languages. today, customisation is an advanced feature, not the baseline. we talk about lisp as a language, but it behaved more like a writable substrate. the cultural part wasn't romanticism it was just people using what was there without artificial walls
(2003, with some footnotes added later)
I think that Python has inherited a lot (but unfortunately not all) of the artistic qualities of Lisp. So in a sense the artistic dynamic approach of Lisp won given the popularity of Python.
I've been realizing lately that we're already to the point where LLMs can basically take written English language and create the computer code it's describing. I think in the future perhaps most (and eventually all) software can be written basically the way we write method/function documentation today. You describe a function, give it a name, describe it's inputs/output arguments. And with enough specificity there will be a [perfect] function that can be written (multiple of course actually) which accomplishes what's described. So we can create entire applications this way. I guess this is already being called no-code or low-code systems, but I think perhaps as a universal language where instructions MUST be given, perhaps LISP is indeed the best because anyone can learn the syntax in basically a couple of minutes.
I can't see how, in 100 years (or maybe even just 50) humans will still need more than ONE computer programming language. Or at the most two. We might always need C++ to exist for low level stuff, but we only need ONE higher level language.
Having developed in a wide variety of different languages in my free time, I really think the idea that languages like lisp and smalltalk give a sense of creative freedom is pretty overblown.
I have realised recently that there are much more concrete barriers to creative expression in coding: many of us are carrying around iPhones that you cannot write applications for, even for yourself, without paying a $100 a year fee. All the ways the language makes you feel means little compared to whether or not you can actually write code for your computer. The person who makes the computer can decide if you are allowed to be creative with it. And, as a society, we have decided (or perhaps been implicitly brainwashed?) to not want to be creative with our computers. In such a situation, the distinction between Java's OOP and Lisp's "computational ideas and expression" means very little.
A bit of a "Stallman was right" moment for me.
I used to feel similar when I’d write Python. It was a beautiful language in its primordial days, and you’d feel like a hipster renegade if you could sneak it into an enterprise environment where C#, Java, and C++ were the norms. Even just whipping up a tiny script that didn’t require compilation and could do some stuff felt like magic.
These days I mostly write Go for work, and as I’ve gotten older, I no longer find the act of programming profound. I take more joy in hiking and kayaking than programming, since LLMs have commoditized something that used to be gatekept to the teeth.
I’m glad that AI tools have trivialized many parts of the act and let people focus on the result rather than the process. Kind of like how good linters completely killed off the bike shedding around code aesthetics.
That said, nowadays I appreciate tools that last. Languages that take backward compatibility seriously and don’t break user code on a whim. Languages that don’t support ten ways of doing the same thing. Languages that don’t require any external dependency managers or build tools. Languages that are fast, have less syntactic noise, and let me do my stuff without much fuss. So to my eyes, those useful languages are the most beautiful.
So Python, with its bolted-on type system, no built-in dependency manager (uv doesn’t count; there will be more unless they put it in the standard toolchain), and a terrible type checker, doesn’t really appeal to me anymore.
I’m sure anyone could write a beautiful ode to any language of their choice and make something profound out of it. If I could, I’d probably write an ode to Go.